CN103577475B - A kind of picture mechanized classification method, image processing method and its device - Google Patents
A kind of picture mechanized classification method, image processing method and its device Download PDFInfo
- Publication number
- CN103577475B CN103577475B CN201210276320.0A CN201210276320A CN103577475B CN 103577475 B CN103577475 B CN 103577475B CN 201210276320 A CN201210276320 A CN 201210276320A CN 103577475 B CN103577475 B CN 103577475B
- Authority
- CN
- China
- Prior art keywords
- picture
- classification
- feature
- characteristic
- sorted
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Active
Links
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F16/00—Information retrieval; Database structures therefor; File system structures therefor
- G06F16/50—Information retrieval; Database structures therefor; File system structures therefor of still image data
- G06F16/58—Retrieval characterised by using metadata, e.g. metadata not derived from the content or metadata generated manually
- G06F16/583—Retrieval characterised by using metadata, e.g. metadata not derived from the content or metadata generated manually using metadata automatically derived from the content
- G06F16/5846—Retrieval characterised by using metadata, e.g. metadata not derived from the content or metadata generated manually using metadata automatically derived from the content using extracted text
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F16/00—Information retrieval; Database structures therefor; File system structures therefor
- G06F16/50—Information retrieval; Database structures therefor; File system structures therefor of still image data
- G06F16/58—Retrieval characterised by using metadata, e.g. metadata not derived from the content or metadata generated manually
- G06F16/583—Retrieval characterised by using metadata, e.g. metadata not derived from the content or metadata generated manually using metadata automatically derived from the content
- G06F16/5838—Retrieval characterised by using metadata, e.g. metadata not derived from the content or metadata generated manually using metadata automatically derived from the content using colour
Landscapes
- Engineering & Computer Science (AREA)
- Library & Information Science (AREA)
- Theoretical Computer Science (AREA)
- Data Mining & Analysis (AREA)
- Databases & Information Systems (AREA)
- Physics & Mathematics (AREA)
- General Engineering & Computer Science (AREA)
- General Physics & Mathematics (AREA)
- Image Analysis (AREA)
- Information Retrieval, Db Structures And Fs Structures Therefor (AREA)
Abstract
The embodiment of the present application discloses a kind of picture mechanized classification method.This method includes:Receive picture to be sorted;Read the feature classification in feature database;The characteristic of the picture to be sorted is extracted according to the feature classification;The characteristic default characteristic corresponding with the feature classification of extraction is matched, the picture merger to be sorted that characteristic can be matched is one kind.The embodiment of the present application additionally provides the image processing method based on picture classification, and picture mechanized classification device and the picture processing device based on picture classification.The embodiment of the present application improves efficiency and the degree of accuracy of picture classification.
Description
Technical field
The application is related to picture Processing Technique field, more particularly to a kind of mass picture mechanized classification method, based on figure
The image processing method of piece classification, and each self-corresponding device.
Background technology
With the development of Internet technology, the increase of information magnanimity, network picture and image data also rapidly increase.Picture is made
For a kind of important information source (carrier), it usually needs carry out centrally stored, processing and some personalized application.But this
Often source is different, form is different for a little pictures, and the difference of graphic form thousand is outer other.To meet the normal utilization to picture, it is often necessary to
These mass pictures are classified, to carry out differential management (processing).
Picture classification mode of the prior art is to use artificial cognition mechanism, i.e., by hand labor by substantial amounts of picture
Class discrimination and merger are carried out one by one according to preset rules.Although this mode is succinctly, conveniently, however, it is desirable to waste a large amount of people
Power, material resources and financial cost, classification effectiveness is relatively low, especially in face of mass picture when, it is even more so.Moreover, artificial cognition side
Formula is due to constantly repeating mechanical work, and easily fatigue, causes classification inaccurate, so as to meet respective practical application needs.
The content of the invention
The problem of existing in view of prior art, the embodiment of the present application provide a kind of picture mechanized classification method, base
Image processing method and its corresponding device in picture classification, to realize picture mechanized classification and processing, improve picture point
The efficiency and the degree of accuracy of class and processing.
The picture mechanized classification method that the embodiment of the present application provides includes:
Receive picture to be sorted;
Read the feature classification in feature database;
The characteristic of the picture to be sorted is extracted according to the feature classification;
The characteristic default characteristic corresponding with the feature classification of extraction is matched, characteristic can be matched
Picture merger to be sorted is one kind.
Preferably, the feature classification includes picture background, picture prospect, picture character and/or picture personage.
It is further preferred that the feature classification is picture background and/or picture prospect, it is described corresponding with feature classification
Default characteristic includes background color number, the background colour of picture background, and/or, the foreground color number of picture prospect, foreground zone
Domain number, foreground location region, then:The characteristic that the picture to be sorted is extracted according to feature classification includes:
The picture to be sorted is divided into one or more subgraphs;
Background color number, the background colour of each subgraph are extracted, and/or, extract foreground color number, the foreground area of each subgraph
Number, foreground area position.
It is further preferred that the extraction foreground color number of subgraph, foreground area number, foreground area position specifically include:
Count pixel number of each color in pre-set color set in the range of the subgraph;
The notable angle value of each color is calculated according to the following formula:
In formula:ciFor i-th of element in pre-set color set, N is the element number of color set, wiFor pre-set color
Pixel number of i-th kind of color in the range of subgraph in set, dist (ck·ci) between two kinds of colors in color set
Distance;
Notable angle value using the notable angle value of color of each pixel in subgraph as the pixel;
It is foreground area by the region recognition that the notable angle value sum of pixel is more than predetermined threshold value;
Foreground color number, the number of foreground area, foreground area position are determined according to the foreground area of identification.
Preferably, the feature classification is picture character, and the default characteristic corresponding with feature classification includes text
Word regional location, character area size and/or word content, then:It is described that the picture to be sorted is extracted according to feature classification
Characteristic includes:
Character area positioning is carried out to the picture to be sorted, to determine character area position and/or character area size;
And/or
Text region is carried out after carrying out character area positioning to the picture to be sorted, to determine in the word of character area
Hold.
Specifically included it is further preferred that carrying out character area positioning to the picture to be sorted:By being united based on texture
Meter method carries out character area positioning based on region analysis method, and/or,
Text region is carried out to the picture to be sorted to specifically include:By KD trees method of identification to the word in character area
It is identified.
Preferably, the feature classification is picture personage, and the default characteristic corresponding with feature classification includes people
Face number, face location and/or face complexion, then:The feature that the picture to be sorted is extracted according to feature classification is to carry
Take face number, face location and/or the face complexion of the picture to be sorted.
The embodiment of the present application additionally provides a kind of image processing method based on picture classification.This method includes:
Receive picture to be sorted;
Read the feature classification in feature database;
The characteristic of the picture to be sorted is extracted according to the feature classification;
The characteristic default characteristic corresponding with the feature classification of extraction is matched, characteristic can be matched
Picture merger to be sorted is one kind;
The picture of one classification is handled according to default processing rule.
Preferably, the default processing rule has corresponding relation with the feature classification in feature database, then described to one
The picture of classification is handled specially according to default processing rule:
Processing rule progress is preset according to the feature classification with being read from feature database is corresponding to the picture of a classification
Processing.
Preferably, the picture to a classification is handled specially according to default processing rule:
Presupposed information is added in the predeterminated position of the picture of a classification.
The embodiment of the present application additionally provides a kind of picture mechanized classification device.The device includes:Receiving unit, read list
Member, extraction unit, matching unit and classification unit, wherein:
The receiving unit, for receiving picture to be sorted;
The reading unit, for reading the feature classification in feature database;
The extraction unit, for extracting the characteristic of the picture to be sorted according to the feature classification of reading;
The matching unit, the picture feature data default characteristic corresponding with the feature classification for matching extraction
According to;
The classification unit, the picture merger to be sorted for characteristic can be matched are one kind.
Preferably, the feature classification is picture background and/or picture prospect, the characteristic corresponding with feature classification
According to background color number, background colour including picture background, and/or, it is the foreground color number of picture prospect, foreground area number, preceding
Scene area position, then the extraction unit include:Subelement and extraction subelement are divided, wherein:
The division unit, for the picture to be sorted to be divided into one or more subgraphs;
The extraction subelement, for extracting background color number, the back of the body of each subgraph according to the feature classification of reading
Scenery, and/or, extract the foreground color number of each subgraph, foreground area number, foreground area position.
It is further preferred that the extraction subelement is used to extract the foreground color number of subgraph, foreground area number, prospect
During regional location, the subelement includes:Pixel statistics subelement, notable angle value computation subunit, foreground area identification are single
Member and determination subelement, wherein:
The pixel counts subelement, for counting each color in pre-set color set in the range of the subgraph
Pixel number;
The significantly angle value computation subunit, for calculating the notable angle value of each color according to the following formula, and by subgraph
Notable angle value of the notable angle value of color of each pixel as the pixel:
In formula:ciFor i-th of element in pre-set color set, N is the element number of color set, wiFor pre-set color
Pixel number of i-th kind of color in the range of subgraph in set, dist (ck·ci) between two kinds of colors in color set
Distance;
The foreground area identifies subelement, for the notable angle value sum of pixel to be more than to the region recognition of predetermined threshold value
For foreground area;
The determination subelement, for determining foreground color number, foreground area number, prospect according to the foreground area of identification
Regional location.
Preferably, the feature classification is picture character, and the default characteristic corresponding with feature classification includes text
Word regional location, character area size and/or word content, then the extraction unit include:Locator unit, and/or, identification
Subelement, wherein:
The locator unit, for carrying out character area positioning to the picture to be sorted, to determine character area position
Put and/or character area size;
The identification subelement, for carrying out Text region after carrying out character area positioning to the picture to be sorted, with
Determine the word content of character area.
It is further preferred that the locator unit based on texture statistics method or based on region analysis method by carrying out word
Zone location, and/or, the word in character area is identified by KD trees method of identification for the identification subelement.
The embodiment of the present application additionally provides a kind of picture processing device based on picture classification.The device includes:Receive single
Member, reading unit, extraction unit, matching unit, classification unit and processing unit, wherein:
The receiving unit, for receiving picture to be sorted;
The reading unit, for reading the feature classification in feature database;
The extraction unit, for extracting the characteristic of the picture to be sorted according to the feature classification of reading;
The matching unit, the picture feature data default characteristic corresponding with the feature classification for matching extraction
According to;
The classification unit, the picture merger to be sorted for characteristic can be matched are one kind;
The processing unit, for being handled according to preset rules the picture of a classification.
The embodiment of the present application extracts picture to be sorted after picture to be sorted is received, according to the feature classification in feature database
Characteristic, then these characteristics default characteristic corresponding with feature classification is matched, if it is possible to
Match somebody with somebody, be then one kind by its merger.Compared with prior art, the embodiment of the present application can obtain automatically feature classification and with feature classification
Corresponding characteristic, and the characteristic of acquisition is matched to distinguish classification with the characteristic extracted, so as to
The classification of mass picture is automatically completed, has saved human and material resources and financial cost, improves the operating efficiency of picture classification.
It is additionally, since and takes automatic business processing measure, as long as pre-sets feature classification and characteristic, you can by matching one by one
Picture classification is realized, so as to improve the accuracy of picture classification.The picture of the embodiment of the present application high efficiency and high accuracy point
Class mode preferably meets the application demand of various occasions.
Brief description of the drawings
, below will be to embodiment or existing in order to illustrate more clearly of the embodiment of the present application or technical scheme of the prior art
There is the required accompanying drawing used in technology description to be briefly described, it should be apparent that, drawings in the following description are only this
Some embodiments described in application, for those of ordinary skill in the art, on the premise of not paying creative work,
Other accompanying drawings can also be obtained according to these accompanying drawings.
Fig. 1 is the picture mechanized classification method flow diagram of the embodiment of the present application one;
Fig. 2 is the picture mechanized classification method flow diagram of the embodiment of the present application two;
Fig. 3 (a)~(f) is the effect diagram of embodiment described in Fig. 2;
Fig. 4 is the structured flowchart of the picture mechanized classification device of the embodiment of the present application three;
Fig. 5 is the structured flowchart of the picture processing device based on picture classification of the embodiment of the present application four.
Embodiment
In order that those skilled in the art more fully understand the technical scheme in the application, it is real below in conjunction with the application
The accompanying drawing in example is applied, the technical scheme in the embodiment of the present application is clearly and completely described, it is clear that described implementation
Example only some embodiments of the present application, rather than whole embodiments.It is common based on the embodiment in the application, this area
The every other embodiment that technical staff is obtained under the premise of creative work is not made, it should all belong to the application protection
Scope.
It is below in conjunction with the accompanying drawings and specific real to enable the above-mentioned purpose of the application, feature and advantage more obvious understandable
Mode is applied to be described in further detail the application.
Embodiment one
Referring to Fig. 1, the figure shows the flow of the picture automatic classification method of the embodiment of the present application one.The present embodiment bag
Include:
Step S101:Receive picture to be sorted;
Picture to be sorted is the picture source that the embodiment of the present application is ready for classification processing, and the picture source can pass through net
The a sheet by a sheet independent picture or be cached to picture in picture library in advance that network receives, the embodiment of the present application are treated point
The source of class picture is simultaneously not intended to be limited in any.When receiving picture to be sorted, a pictures once can be only received, can also be once
Plurality of pictures is received, this depends on the disposal ability of picture classification device and the requirement to picture classification device operating efficiency.
Step S102:Read the feature classification in feature database;
Feature database is the database for being exclusively used in placing feature classification.It is special corresponding to feature classification in actual application
Sign data can also be stored in advance in this feature storehouse, can also in picture classification processing procedure pop-up dialogue box, it is desirable to use
Characteristic corresponding to the input feature vector classification of family.User can need selection any of which mode, Qian Zheyou according to oneself
In pre-set unified characteristic, thus it is commonly available to large batch of picture and classifies automatically, the latter is due to can be with
When adjust characteristic, thus multi-level, the personalized classification of picture can be realized.
Step S103:The characteristic of the picture to be sorted is extracted according to the feature classification;
After obtaining feature classification, for ease of carrying out the matching of characteristic, it is necessary to be extracted from picture to be sorted correspondingly
Characteristic.Different feature classifications, the extracting method of its characteristic there may be difference, such as the background of extraction picture
Data can use picture cutting technique and picture marking area extractive technique, and picture character can be used by extracting the lteral data of picture
Positioning and character recognition technology.
Step S104:The characteristic default characteristic corresponding with the feature classification of extraction is matched, by characteristic
It is one kind according to the picture merger to be sorted that can be matched.
After the characteristic that picture to be sorted is extracted by preceding feature extraction step, you can by the data and feature class
Preset characteristic corresponding to mesh to be matched, " default characteristic corresponding with feature classification " here, as it was previously stated, can
Being stored in advance in feature database, setting read when needed from feature database or interim.Matching operation is specific
Can disposably be matched after a number of picture to be sorted of extraction or a feature extraction one during progress
Secondary matching process.By matching, picture to be sorted can be divided at least two classes:Picture to be sorted that characteristic matches and
The unmatched picture to be sorted of characteristic.When feature classification is stagewise feature classification, the classification of division can be multistage.
How much relevant with picture source and feature classification and corresponding default characteristic the picture number of one kind is integrated into, in picture source
When fixed, if user needs picture carrying out multiscale category division, feature classification can be refined, preset multiple layers
Secondary, multiple ranks;If user is higher to category images accuracy requirement, the standard of characteristic can be improved, or by feature
Data are arranged to " section ", rather than single " numerical value ".These different implementations depend on the various tools of practical application
Body demand.
The present embodiment extracts the spy of picture to be sorted according to the feature classification in feature database after picture to be sorted is received
Data are levied, are then matched these characteristics default characteristic corresponding with feature classification, if it is possible to match, then
It is one kind by its merger.Compared with prior art, the present embodiment can obtain feature classification and spy corresponding with feature classification automatically
Data are levied, and the characteristic of acquisition is matched to distinguish classification with the characteristic extracted, so as to automatically complete
Into the classification of mass picture, human and material resources and financial cost have been saved, have improved the operating efficiency of picture classification.It is additionally, since
Automatic business processing measure is taken, as long as pre-setting feature classification and characteristic, you can realize picture by matching one by one
Classification, so as to improve the accuracy of picture classification.The picture classification mode of the embodiment of the present application high efficiency and high accuracy compared with
The application demand of various occasions is met well.In addition, according to the various concrete scenes of practical application, multiple features classification can be passed through
Multi-level picture classification is realized, is set by characteristic high standard, range selector and realizes high-precision picture classification.
Embodiment two
To further illustrate the technical scheme of the application, illustrated below with a more specific example.In the reality
Apply in example, feature classification includes:Picture background, picture prospect, picture character and picture personage, picture background feature classification are corresponding
Default characteristic be background color number, background colour, the characteristic of picture prospect is foreground object number, foreground object position
Put, characteristic corresponding to picture character is character area position, character area size and word content, corresponding to picture personage
Characteristic face number, face location and face complexion.Characteristic corresponding to these feature classifications and its feature classification
Value has been stored in advance among feature database.
Referring to Fig. 2, the figure shows the flow of the picture mechanized classification method of the present embodiment two.The embodiment includes:
Step S201:Four picture background, picture prospect in reading feature database, picture character and picture personage feature classes
Mesh and its corresponding characteristic;
It is because the present embodiment is adopted that the operation for reading feature classification and characteristic is placed on the first step and performed by the present embodiment
With the characteristic being stored in advance in feature database, it is not necessary to interim as needed in picture classification processing procedure to set feature
Data.
Step S202:Receive a picture to be sorted;
Step S203:Two background complexity, background colour background characteristics of the picture to be sorted are extracted, extract institute
The prospect complexity, foreground area number, the foreground features data of foreground area position three of picture to be sorted are stated, extraction is to be sorted
Three the character area position of picture, character area size and word content character features data, and extraction picture to be sorted
Face number, three character features data of face location and face complexion;
As it was previously stated, the method for extracting the characteristic of this feature classification of to be sorted picture different according to feature classification
May be different.For example for extracting the characteristic of picture foreground and background featured items, first it can use picture segmentation technology will
Picture is divided into multiple subgraphs, then takes the Color Color number and foreground area of picture marking area extractive technique extraction picture
Number, foreground area position, on this technical foundation, also extractable picture background color etc.;For extracting picture character feature
The characteristic of project, character area position and character area size can first be determined using picture character area positioning technology,
On this technical foundation, word content is identified using character recognition technology;, can for the characteristic of picture character features classification
To advanced row Face datection, Face Detection is carried out again after detecting face, to determine characteristic.This implementation is introduced in turn below
The specific method of the preferable extraction characteristic of example:
For the picture segmentation method by picture segmentation into subgraph:First calculate the gray scale of each row (column) neighbor pixel of picture
Gradient difference value, sharp point being judged according to pixel gray level variation tendency, the line segment then formed to boundary point is analyzed,
Judge whether these boundary sections form the specific region of the condition of satisfaction, if it is, determine that the region is image boundary, and then
Picture segmentation is carried out according to the specific region, the mode that straight line intersects in length and breadth can be taken during segmentation to be split, be so advantageous to
Computerization processing is carried out, can also take other modes to split.It is worth noting that, work as a width picture only and include an image
When, the border searched out is picture boundary, and when a picture includes multiple image, multiple figures can be obtained by the above method
As region, these image-regions form " subgraph " of picture.
For the extracting method of picture marking area, including:
(1) pixel number of each color in the range of the subgraph in pre-set color set, pre-set color collection are counted
Conjunction is a preassigned color space, for example can be Lab space, the color elements that color set includes set C=
{c1, c2..., cNRepresent, the element number of the set is N value typically hundreds of kinds of colors;
(2) the notable angle value of each color is calculated according to the following formula:
In formula:ciFor i-th of element in pre-set color set, N is the element number of color set, wiFor pre-set color
Pixel number of i-th kind of color in the range of subgraph in set, dist (ck·ci) between two kinds of colors in color set
Distance;
(3) using the notable angle value of color of each pixel in subgraph as the notable angle value of the pixel, pixel is shown
The region recognition that angle value sum is write more than predetermined threshold value is foreground area;
(4) foreground color number, foreground area position, prospect can thus be known after determining foreground area according to above-mentioned steps
The information such as areal.
For the localization method of character area in picture, the present embodiment provides the specific method for optimizing of the following two kinds:
First, being based on texture statistics method, this method catches word to have unique texture features, such as word has strong side
Edge, word region gray-value variation are larger, background color before word region has substantially, more than word region
There is blank with following part, judges whether certain region is character area by grader using these features, its step bag
Include:(1) with the sliding window scan image of different scale to obtain multiple candidate windows;(2) each candidate window is directed to, is calculated
The textural characteristics of pixel in the candidate window, textural characteristics can by calculate gray scale difference value, extraction edge feature, carry out it is discrete
The modes such as cosine transform (DCT), wavelet transformation, Gabor transformation obtain;(3) by grader judge each candidate window whether be
Character area, and then realize the positioning of character area.
Second, being based on region analysis method, this method catches word to have obvious provincial characteristics, for example most of word is all
Closing, independent region, either because of the close formation connected component of pixel color or by a strong edge delineation in an envelope
In closed region, its step includes:(1) connected component all in image is extracted, extraction process can use binarization of gray value, color
The methods of cluster, rim detection, is realized;(2) non-legible region, these characteristic informations are filtered using the characteristic information of connected component
Including geological information (such as area, length-width ratio, numbers of hole), fitted ellipse information, stroke information, bone information, solid colour
Property etc.;(3) discrete word connected component is merged using bottom-up method, determines the border of character area, from
And realize character area positioning and obtain character area size.
After the character area in image is obtained by abovementioned steps, it can realize that word carries further directed to the character area
Take, Word Input can use KD trees method of identification to realize, this method includes:Set KD trees identification rudimentary model;Travel through character area
Interior word, the word identified for the first time is trained, to correct KD tree identification models, by the model again identify that directly
To the whole words identified in character area.In order to improve recognition effect, then character area can also be carried out before identifying
Pretreatment, pretreatment include typesetting analysis, binaryzation, text line detect, cut word etc..
For method for detecting human face, the present embodiment is preferably carried out in accordance with the following steps:First by manually marking substantial amounts of face
Picture and non-face picture, quasi-Haar wavelet feature extraction is carried out for these faces and non-face picture, then according to extraction
Quasi-Haar wavelet characteristic use Adboost learn differentiation face and non-face hierarchical classification device;Slided with different scale
Window scan image obtains candidate's subwindow;Differentiate whether each candidate window is face using hierarchical classification device, finally to judging
The candidate window of face merges when out, obtains final human face region.Image is may determine therefrom that after determining human face region
In face number, face location and face size.For Face Detection, the present embodiment is preferably carried out in accordance with the following steps:First
By manually marking substantial amounts of colour of skin picture, learn the gauss hybrid models of skin color for these colour of skin pictures;According to skin
Color gauss hybrid models calculate probability that each pixel in image is the colour of skin to obtain skin-color probability distributions figure, by the general of pixel
For rate compared with given threshold, the region more than the threshold value is the area of skin color with the specific colour of skin.
Step S204:Judge whether the characteristic characteristic corresponding with the feature classification of extraction matches, if
Matching, then perform step S205;If it does not match, perform step S206;
Here the matching between the characteristic extracted and default characteristic may be specific for different characteristics
Mode has differences, such as, for the characteristic of background complexity, it is that numeric type compares, if complexity for 100 points
System, as long as then the complexity value operation pre-set threshold value (default characteristic) of the characteristic of extraction is regarded to realization
Match somebody with somebody;For the characteristic of background colour, it is that two-valued logic type compares, if default characteristic background colour is white pure color,
Then as long as the background colour of the characteristic of extraction is thought to realize for its allochromatic colour (including non-white pure color, tertiary colour etc.)
Matching, the matching process of other characteristics are similar.
Step S205:It is one kind by picture merger to be sorted;These pictures for being classified as one kind can be stored, so as to follow-up
Program operates on it.
Step S206:Judge whether picture to be sorted is disposed, if it is not, then return to step S202;If it is, knot
Beam picture classification flow.
The present embodiment is according to four picture background, picture prospect, picture character and picture personage feature classifications to be sorted
Picture has carried out mechanized classification.Fig. 3 shows the design sketch of the present embodiment, wherein:Fig. 3 (a) is the artwork of picture to be sorted;
Fig. 3 (b) is characterized project and characteristic, and the characteristic background colour for listing picture background feature classification herein is pure for white
Color background, picture foreground location are picture centre position, and the characteristic of picture character feature classification is commodity LOGO
(O.S.A) it is, to have model in the upper left corner, the characteristic of picture character features classification;Fig. 3 (c) is the marking area of extraction
Figure, translucent grid area is background;Fig. 3 (d) is the character area figure of extraction, and translucent yellow area is character area;Fig. 3
(e) it is Face detection figure, the region in blue box is face.
It is worth noting that, multiple feature classifications are have read simultaneously when the present embodiment reads feature classification, and in feature
Also multiple characteristics are set in the setting of the characteristic of classification, in fact, in industry application, can be by these feature classifications
Independently carried out with characteristic, or even on the basis of the present embodiment, increase more feature classifications and characteristic.These
Selection under various situations each depends on the degree of accuracy to picture classification and the requirement of efficiency, generally, feature classification and characteristic
According to more, the degree of accuracy of picture classification is higher, but due to extracting and the increase of the workload of matching process, the efficiency of picture classification is again
It will be affected, therefore, in terms of comprehensive, it is necessary to which the needs for balancing these two aspects come actual selected characteristic classification and characteristic.
Further, the present embodiment has carried out mechanized classification to picture.On the basis of picture classification, it can carry out
The processing of picture, such as, the processing such as unitized mark superposition, picture cutting, picture pixels adjustment are carried out on the picture of classification.
Specifically carry out which kind of type of process and how to handle to be set by default processing rule, such as, it is now desired to it is right
The size dimension of the picture of one classification carries out unification, can setting processing type be " cuttings ", handling regular can set cutting
Width, length etc..In a particular application, default processing rule can also be connected with feature classification, i.e., certain feature
Classification corresponds to certain default processing rule, such as, feature classification if picture background, then preset processing rule can be by
The picture background of the picture of one classification is strengthened or decrease processing, to adapt to the needs of certain light intensity, also for example, feature class
Mesh is picture character, then it can be the adjustment for entering line position sequence, size, color etc. to the word in picture to preset processing rule.
Exemplified by stamping unified mark in the precalculated position of picture:Because picture is classified according to pre-provisioning request,
So that the operation of unified marking becomes simple, quick, it is thus only necessary to it is mechanical to repeat marking, without worrying the behaviour
Whether other picture elements are hindered.Accordingly, with respect to non-classified mass picture, the efficiency of present treatment process can carry significantly
It is high.It can facilitate by the picture of unified franking and picture is shown, managed, especially in commercial operation platform, to business
The unitized displaying of product (such as clothes, knapsack), brand names image can be strengthened, improve the clicking rate and buying rate of commodity.Ginseng
See Fig. 3 (f), the figure shows the effect after mark (Tmall.com) is stamped by the lower right corner of sorted picture.
Embodiment three
The present processes are described in detail in above example, and correspondingly, present invention also provides a kind of picture to divide automatically
Class device.Referring to Fig. 4, the figure shows the structured flowchart of the picture mechanized classification device of embodiments herein three.This dress
Putting embodiment 400 includes:Receiving unit 401, reading unit 402, extraction unit 403, matching unit 404 and classification unit 405,
Wherein:
Receiving unit 401, for receiving picture to be sorted;
Reading unit 402, for reading the feature classification in feature database;
Extraction unit 403, for extracting the characteristic of the picture to be sorted according to the feature classification of reading;
Matching unit 404, the picture feature data default characteristic corresponding with the feature classification for matching extraction
According to;
Sort out unit 405, the picture merger to be sorted for characteristic can be matched is one kind.
The course of work of present apparatus embodiment 400 is:Receiving unit 401 receives picture to be sorted and reading unit 402 is read
After taking the feature classification in feature database, the picture to be sorted is extracted according to the feature classification of reading by extraction unit 403
Characteristic, then, matching unit 404 is by the picture feature data of extraction default characteristic corresponding with the feature classification
According to matching operation is carried out, after overmatching, the picture merger to be sorted that sorting out unit 405 can match characteristic is one kind.
Present apparatus embodiment is receiving picture to be sorted and is extracting picture to be sorted according to the feature classification in feature database
Characteristic after, these characteristics default characteristic corresponding with feature classification is matched, if it is possible to match,
It is then one kind by its merger.Compared with prior art, present apparatus embodiment can obtain automatically feature classification and with feature classification pair
The characteristic answered, and the characteristic of acquisition is matched to distinguish classification with the characteristic extracted, so as to certainly
The classification of mass picture is completed dynamicly, has saved human and material resources and financial cost, improves the operating efficiency of picture classification.And
And due to taking automatic business processing measure, as long as pre-setting feature classification and characteristic, you can by matching reality one by one
Existing picture classification, so as to improve the accuracy of picture classification.The picture classification of present apparatus embodiment high efficiency and high accuracy
Mode preferably meets the application demand of various occasions.
It is worth noting that, the receiving unit 401 and reading unit 402 of said apparatus embodiment are according to the need of practical application
Will, the operation that can first carry out receiving unit 401 performs the operation of reading unit 402 again, also can be on the contrary, even both hold simultaneously
OK.Receiving unit first carries out operation and is primarily adapted for use in the situation for needing that personalized classification is carried out for different pictures to be sorted,
So, user can set different feature classifications according to the first impression for treating category images to a number of picture to be sorted
And characteristic, this is adapted to carry out a small amount of picture " essence is classified ";Reading unit, which first carries out operation and is primarily adapted for use in, needs pin
The situation of homogeneous classification is carried out to the picture to be sorted of magnanimity, so only needs to perform a read operation, follow-up each treats
Category images uses this feature classification and characteristic, and this is suitable for carrying out mass picture " just classification ".The present apparatus is answered
User can be according to the various execution sequences that above-mentioned associative cell is selected with application scenarios.
Extraction unit 403 in said apparatus embodiment can have in the case of the feature classification data difference of processing
There is different internal structure.In order to avoid repeating, the technique effect brought below for extraction unit different internal structure can be found in
The description of aforementioned manner embodiment appropriate section.
(1) when feature classification is picture background and/or picture prospect, the characteristic corresponding with feature classification includes
Background color number, the background colour of picture background, and/or, the foreground color number of picture prospect, foreground area number, foreground area
During position, extraction unit 403 can include:Subelement and extraction subelement are divided, wherein:Division unit, for being treated described
Category images is divided into one or more subgraphs;Subelement is extracted, for extracting each height according to the feature classification of reading
Background color number, the background colour of figure, and/or, extract the foreground color number of each subgraph, foreground area number, foreground area position.
It is further preferred that when extraction subelement is used to extract the foreground color number of subgraph, foreground area number, foreground zone
During the position of domain, the subelement can also include:Pixel statistics subelement, notable angle value computation subunit, foreground area identification
Subelement and determination subelement, wherein:Pixel counts subelement, for counting each color in pre-set color set in institute
State the pixel number in the range of subgraph;The significantly angle value computation subunit, for calculating the notable of each color according to the following formula
Angle value, and the notable angle value using the notable angle value of color of each pixel in subgraph as the pixel:
In formula:ciFor i-th of element in pre-set color set, N is the element number of color set, wiFor pre-set color
Pixel number of i-th kind of color in the range of subgraph in set, dist (ck·ci) between two kinds of colors in color set
Distance;
Foreground area identifies subelement, before the region recognition for the notable angle value sum of pixel to be more than to predetermined threshold value is
Scene area;Determination subelement, for determining foreground color number, foreground area number, foreground area according to the foreground area of identification
Position.
(2) when feature classification is picture character, the characteristic corresponding with feature classification include character area position,
When character area size and/or word content, extraction unit 403 can include:Locator unit, and/or, subelement is identified,
Wherein:Locator unit, for carrying out character area positioning to the picture to be sorted, with determine character area position and/or
Character area size;Subelement is identified, for carrying out Text region after carrying out character area positioning to the picture to be sorted, with
Determine the word content of character area.It is further preferred that locator unit passes through based on texture statistics method or based on region point
Analysis method carries out character area positioning, and/or, the identification subelement is carried out by KD trees method of identification to the word in character area
Identification.
Example IV
Embodiment three realizes the mechanized classification to picture.In most cases, picture is classified is not most
Whole purpose, also need to carry out picture processing on basis of classification.Picture processing includes various types of other processing, such as, classifying
Picture on carry out unitized mark superposition, picture is cut, picture pixels adjustment etc..The present embodiment put forward one kind in classification chart
The device of pictorial information addition is carried out on piece basis.Referring to Fig. 5, the figure shows the embodiment of the present application four based on picture classification
Picture processing device composition frame chart.Present apparatus embodiment 500 includes:Receiving unit 501, reading unit 502, extraction unit
503rd, matching unit 504, classification unit 505 and processing unit 506, wherein:
Receiving unit 501, for receiving picture to be sorted;
Reading unit 502, for reading the feature classification in feature database;
Extraction unit 503, for extracting the characteristic of the picture to be sorted according to the feature classification of reading;
Matching unit 504, the picture feature data default characteristic corresponding with the feature classification for matching extraction
According to;
Sort out unit 505, the picture merger to be sorted for characteristic can be matched is one kind;
Processing unit 506, the predeterminated position for the picture in a classification add presupposed information.
The course of work of present apparatus embodiment 500 is:Receiving unit 501 receives picture to be sorted and reading unit 502 is read
After taking the feature classification in feature database, the picture to be sorted is extracted according to the feature classification of reading by extraction unit 503
Characteristic, then, matching unit 504 is by the picture feature data of extraction default characteristic corresponding with the feature classification
According to matching operation is carried out, after overmatching, the picture merger to be sorted that sorting out unit 505 can match characteristic is one kind,
Then presupposed information is added by predeterminated position of the processing unit in the picture of a classification.
Present apparatus embodiment is carrying out before processing to picture, due to taking picture classification measure so that the figure of processing
Piece has relatively uniform form, size etc., forms internal requirement feature, locates rapidly so as to be advantageous to carry out the batch of picture
Reason, improve the treatment effeciency of picture.
It is worth noting that, in the present embodiment in addition to picture processing unit, other units can be integrated into picture classification device, enter
And the various concrete applications based on the picture classification device are established, and picture classification device can also have the various tools of embodiment three
Body structure, to meet the processing requirement to different pieces of information object.For example commercialized running platform needs a large amount of clothes to shooting
Picture be shown on website.Due to shooting clothes factor (such as clothes classification, color, size etc.) in itself and
The influence of each side conditions such as photographed scene (such as daytime, evening etc.), clap the picture that color comes out and differ larger, at this moment can use
The picture classification device that the application provides carries out the classification of default feature classification and characteristic to these pictures, then passes through exhibition again
Show that unit will come out by the sorted picture presentation of picture classification device., can also be by processing unit same before displaying
Commodity LOGO information is added on the predeterminated position of the picture of classification.Picture presentation efficiency is so not only increased, and is enhanced
The brand image of enterprise.
Each embodiment in this specification is described by the way of progressive, identical similar portion between each embodiment
Divide mutually referring to what each embodiment stressed is the difference with other embodiment.It is real especially for system
For applying example, because it is substantially similar to embodiment of the method, so describing fairly simple, related part is referring to embodiment of the method
Part explanation.System embodiment described above is only schematical, wherein described be used as separating component explanation
Unit can be or may not be physically separate, can be as the part that unit is shown or may not be
Physical location, you can with positioned at a place, or can also be distributed on multiple NEs.Can be according to the actual needs
Some or all of module therein is selected to realize the purpose of this embodiment scheme.Those of ordinary skill in the art are not paying
In the case of creative work, you can to understand and implement.
The application can be used in numerous general or special purpose computing system environments or configuration.Such as:Personal computer, service
Device computer, handheld device or portable set, laptop device, multicomputer system, the system based on microprocessor, top set
Box, programmable consumer-elcetronics devices, network PC, minicom, mainframe computer including any of the above system or equipment
DCE etc..
The application can be described in the general context of computer executable instructions, such as program
Module.Usually, program module includes performing particular task or realizes routine, program, object, the group of particular abstract data type
Part, data structure etc..The application can also be put into practice in a distributed computing environment, in these DCEs, by
Task is performed and connected remote processing devices by communication network.In a distributed computing environment, program module can be with
In the local and remote computer-readable storage medium including storage device.
Described above is only the embodiment of the application, it is noted that for the ordinary skill people of the art
For member, on the premise of the application principle is not departed from, some improvements and modifications can also be made, these improvements and modifications also should
It is considered as the protection domain of the application.
Claims (16)
- A kind of 1. picture mechanized classification method, it is characterised in that this method includes:Receive picture to be sorted;Read the feature classification in feature database;The characteristic of the picture to be sorted is extracted according to the feature classification;Match extraction characteristic it is corresponding with the feature classification preset characteristic, treated what characteristic can match point The merger of class picture is one kind.
- 2. according to the method for claim 1, it is characterised in that the feature classification includes picture background, picture prospect, figure Piece word and/or picture personage.
- 3. according to the method for claim 2, it is characterised in that the feature classification is picture background and/or picture prospect, The default characteristic corresponding with feature classification includes background color number, the background colour of picture background, and/or, before picture The foreground color number of scape, foreground area number, foreground area position, then:It is described that the figure to be sorted is extracted according to feature classification The characteristic of piece includes:The picture to be sorted is divided into one or more subgraphs;Background color number, the background colour of each subgraph are extracted, and/or, foreground color number, the foreground area for extracting each subgraph are individual Number, foreground area position.
- 4. according to the method for claim 3, it is characterised in that extract the foreground color number of subgraph, foreground area number, preceding Scene area position specifically includes:Count pixel number of each color in pre-set color set in the range of the subgraph;The notable angle value of each color is calculated according to the following formula:<mrow> <mi>s</mi> <mrow> <mo>(</mo> <msub> <mi>c</mi> <mi>k</mi> </msub> <mo>)</mo> </mrow> <mo>=</mo> <munderover> <mi>&Sigma;</mi> <mrow> <mi>i</mi> <mo>=</mo> <mn>1</mn> </mrow> <mi>N</mi> </munderover> <msub> <mi>w</mi> <mi>i</mi> </msub> <mo>&CenterDot;</mo> <mi>dist</mi> <mrow> <mo>(</mo> <msub> <mi>c</mi> <mi>k</mi> </msub> <mo>&CenterDot;</mo> <msub> <mi>c</mi> <mi>i</mi> </msub> <mo>)</mo> </mrow> </mrow>In formula:ciFor i-th of element in pre-set color set, N is the element number of color set, wiFor pre-set color set Pixel number of the i-th kind of interior color in the range of subgraph, dist (ck·ci) between two kinds of colors in color set away from From;Notable angle value using the notable angle value of color of each pixel in subgraph as the pixel;It is foreground area by the region recognition that the notable angle value sum of pixel is more than predetermined threshold value;Foreground color number, foreground area number, foreground area position are determined according to the foreground area of identification.
- 5. according to the method for claim 2, it is characterised in that the feature classification is picture character, described and feature class Characteristic is preset corresponding to mesh includes character area position, character area size and/or word content, then:It is described according to spy The characteristic that sign classification extracts the picture to be sorted includes:Character area positioning is carried out to the picture to be sorted, to determine character area position and/or character area size;With/ Or,Text region is carried out after carrying out character area positioning to the picture to be sorted, to determine the word content of character area.
- 6. according to the method for claim 5, it is characterised in that it is specific that character area positioning is carried out to the picture to be sorted Including:By carrying out character area positioning based on texture statistics method or based on region analysis method, and/or,Text region is carried out to the picture to be sorted to specifically include:The word in character area is carried out by KD trees method of identification Identification.
- 7. according to the method for claim 2, it is characterised in that the feature classification is picture personage, described and feature class Characteristic is preset corresponding to mesh includes face number, face location and/or face complexion, then:It is described to be carried according to feature classification The characteristic of the picture to be sorted is taken as face number, face location and/or the face skin of the extraction picture to be sorted Color.
- 8. a kind of image processing method based on picture classification, it is characterised in that this method includes:Receive picture to be sorted;Read the feature classification in feature database;The characteristic of the picture to be sorted is extracted according to the feature classification;Match extraction characteristic it is corresponding with the feature classification preset characteristic, treated what characteristic can match point The merger of class picture is one kind;The picture of one classification is handled according to default processing rule.
- 9. according to the method for claim 8, it is characterised in that the default processing rule and the feature classification in feature database With corresponding relation, then the picture to a classification is handled specially according to default processing rule:The picture of one classification is handled according to the corresponding default processing rule of the feature classification with being read from feature database.
- 10. according to the method for claim 8, it is characterised in that the picture to a classification is advised according to default processing Then handled specially:Presupposed information is added in the predeterminated position of the picture of a classification.
- 11. a kind of picture mechanized classification device, it is characterised in that the device includes:Receiving unit, reading unit, extraction are single Member, matching unit and classification unit, wherein:The receiving unit, for receiving picture to be sorted;The reading unit, for reading the feature classification in feature database;The extraction unit, for extracting the characteristic of the picture to be sorted according to the feature classification of reading;The matching unit, the picture feature data default characteristic corresponding with the feature classification for matching extraction;The classification unit, the picture merger to be sorted for characteristic can be matched are one kind.
- 12. device according to claim 11, it is characterised in that before the feature classification is picture background and/or picture Scape, the default characteristic corresponding with feature classification include background color number, the background colour of picture background, and/or, picture The foreground color number of prospect, foreground area number, foreground area position, then the extraction unit include:Divide subelement and carry Subelement is taken, wherein:The division unit, for the picture to be sorted to be divided into one or more subgraphs;The extraction subelement, for extracting background color number, the background colour of each subgraph according to the feature classification of reading, And/or extract the foreground color number of each subgraph, foreground area number, foreground area position.
- 13. device according to claim 12, it is characterised in that the extraction subelement is used for the prospect face for extracting subgraph When chromatic number, foreground area number, foreground area position, the subelement includes:Pixel statistics subelement, notable angle value calculate son Unit, foreground area identification subelement and determination subelement, wherein:The pixel counts subelement, for counting picture of each color in pre-set color set in the range of the subgraph Vegetarian refreshments number;The significantly angle value computation subunit, for calculating the notable angle value of each color according to the following formula, and will be each in subgraph Notable angle value of the notable angle value of color of pixel as the pixel:<mrow> <mi>s</mi> <mrow> <mo>(</mo> <msub> <mi>c</mi> <mi>k</mi> </msub> <mo>)</mo> </mrow> <mo>=</mo> <munderover> <mi>&Sigma;</mi> <mrow> <mi>i</mi> <mo>=</mo> <mn>1</mn> </mrow> <mi>N</mi> </munderover> <msub> <mi>w</mi> <mi>i</mi> </msub> <mo>&CenterDot;</mo> <mi>dist</mi> <mrow> <mo>(</mo> <msub> <mi>c</mi> <mi>k</mi> </msub> <mo>&CenterDot;</mo> <msub> <mi>c</mi> <mi>i</mi> </msub> <mo>)</mo> </mrow> </mrow>In formula:ciFor i-th of element in pre-set color set, N is the element number of color set, wiFor pre-set color set Pixel number of the i-th kind of interior color in the range of subgraph, dist (ck·ci) between two kinds of colors in color set away from From;The foreground area identifies subelement, before the region recognition for the notable angle value sum of pixel to be more than to predetermined threshold value is Scene area;The determination subelement, for determining foreground color number, foreground area number, foreground area according to the foreground area of identification Position.
- 14. device according to claim 11, it is characterised in that the feature classification is picture character, described and feature Characteristic is preset corresponding to classification includes character area position, character area size and/or word content, then the extraction is single Member includes:Locator unit, and/or, subelement is identified, wherein:The locator unit, for carrying out character area positioning to the picture to be sorted, with determine character area position and/ Or character area size;The identification subelement, for carrying out Text region after carrying out character area positioning to the picture to be sorted, to determine The word content of character area.
- 15. device according to claim 14, it is characterised in that the locator unit by based on texture statistics method or Character area positioning is carried out based on region analysis method, and/or, the identification subelement is by KD tree method of identifications in character area Word be identified.
- 16. a kind of picture processing device based on picture classification, it is characterised in that the device includes:Receiving unit, read list Member, extraction unit, matching unit, classification unit and processing unit, wherein:The receiving unit, for receiving picture to be sorted;The reading unit, for reading the feature classification in feature database;The extraction unit, for extracting the characteristic of the picture to be sorted according to the feature classification of reading;The matching unit, the picture feature data default characteristic corresponding with the feature classification for matching extraction;The classification unit, the picture merger to be sorted for characteristic can be matched are one kind;The processing unit, handled for the picture to a classification according to default processing rule.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201210276320.0A CN103577475B (en) | 2012-08-03 | 2012-08-03 | A kind of picture mechanized classification method, image processing method and its device |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201210276320.0A CN103577475B (en) | 2012-08-03 | 2012-08-03 | A kind of picture mechanized classification method, image processing method and its device |
Publications (2)
Publication Number | Publication Date |
---|---|
CN103577475A CN103577475A (en) | 2014-02-12 |
CN103577475B true CN103577475B (en) | 2018-01-30 |
Family
ID=50049273
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN201210276320.0A Active CN103577475B (en) | 2012-08-03 | 2012-08-03 | A kind of picture mechanized classification method, image processing method and its device |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN103577475B (en) |
Families Citing this family (23)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN104915344B (en) * | 2014-03-10 | 2019-04-26 | 联想(北京)有限公司 | A kind of information processing method and electronic equipment |
WO2015139204A1 (en) * | 2014-03-18 | 2015-09-24 | 华为终端有限公司 | Picture management method and device |
CN105608459B (en) * | 2014-10-29 | 2018-09-14 | 阿里巴巴集团控股有限公司 | The dividing method and its device of commodity picture |
CN104331490B (en) * | 2014-11-14 | 2018-07-13 | 北京国双科技有限公司 | network data processing method and device |
CN105843827A (en) * | 2015-05-27 | 2016-08-10 | 维沃移动通信有限公司 | Picture classification method and device for mobile terminal |
CN105120335B (en) * | 2015-08-17 | 2018-08-24 | 无锡天脉聚源传媒科技有限公司 | A kind of method and apparatus of processing TV programme picture |
CN105120334B (en) * | 2015-08-17 | 2018-12-21 | 无锡天脉聚源传媒科技有限公司 | A kind of method and apparatus handling TV programme picture |
CN105245440A (en) * | 2015-10-23 | 2016-01-13 | 努比亚技术有限公司 | Method, system and operation method for pushing picture messages |
CN105868680A (en) * | 2015-11-24 | 2016-08-17 | 乐视致新电子科技(天津)有限公司 | Channel logo classification method and apparatus |
CN106886789A (en) * | 2015-12-16 | 2017-06-23 | 芋头科技(杭州)有限公司 | A kind of image recognition sorter and method |
CN107016392A (en) * | 2016-01-27 | 2017-08-04 | 四川效率源信息安全技术股份有限公司 | A kind of method of text border in removal picture |
CN106021573A (en) * | 2016-05-31 | 2016-10-12 | 北京小米移动软件有限公司 | Image classification method and device and electronic device |
CN106547893A (en) * | 2016-11-03 | 2017-03-29 | 福建中金在线信息科技有限公司 | A kind of photo sort management system and photo sort management method |
CN106682683B (en) * | 2016-11-03 | 2020-09-29 | 知酒(上海)网络科技有限公司 | Wine label picture identification method and device |
CN106599940B (en) * | 2016-11-25 | 2020-04-17 | 东软集团股份有限公司 | Picture character recognition method and device |
CN107066571A (en) * | 2017-04-07 | 2017-08-18 | 深圳市金立通信设备有限公司 | A kind of image processing method and its equipment |
CN107944022A (en) * | 2017-12-11 | 2018-04-20 | 努比亚技术有限公司 | Picture classification method, mobile terminal and computer-readable recording medium |
CN108595661A (en) * | 2018-04-28 | 2018-09-28 | 珠海格力电器股份有限公司 | Method, system, storage medium, photographing device and the mobile phone of photo disposal |
CN108717530B (en) * | 2018-05-21 | 2021-06-25 | Oppo广东移动通信有限公司 | Image processing method, image processing device, computer-readable storage medium and electronic equipment |
CN110097082B (en) * | 2019-03-29 | 2021-08-27 | 广州思德医疗科技有限公司 | Splitting method and device of training set |
CN110222207B (en) * | 2019-05-24 | 2021-03-30 | 珠海格力电器股份有限公司 | Picture sorting method and device and intelligent terminal |
CN110362702B (en) * | 2019-05-27 | 2021-07-20 | 中国联合网络通信集团有限公司 | Picture management method and equipment |
CN116912721B (en) * | 2023-09-14 | 2023-12-05 | 众芯汉创(江苏)科技有限公司 | Power distribution network equipment body identification method and system based on monocular stereoscopic vision |
Family Cites Families (5)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
JP3156667B2 (en) * | 1998-06-01 | 2001-04-16 | 日本電気株式会社 | Digital watermark insertion system, digital watermark characteristic table creation device |
US7031518B2 (en) * | 2002-07-01 | 2006-04-18 | Xerox Corporation | Segmentation method and system for Multiple Raster Content (MRC) representation of documents |
CN101055592A (en) * | 2007-05-20 | 2007-10-17 | 宁尚国 | Image information generation method and device |
CN101635763A (en) * | 2008-07-23 | 2010-01-27 | 深圳富泰宏精密工业有限公司 | Picture classification system and method |
JP4963306B2 (en) * | 2008-09-25 | 2012-06-27 | 楽天株式会社 | Foreground region extraction program, foreground region extraction device, and foreground region extraction method |
-
2012
- 2012-08-03 CN CN201210276320.0A patent/CN103577475B/en active Active
Also Published As
Publication number | Publication date |
---|---|
CN103577475A (en) | 2014-02-12 |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
CN103577475B (en) | A kind of picture mechanized classification method, image processing method and its device | |
Jasiewicz et al. | Landscape similarity, retrieval, and machine mapping of physiographic units | |
Balaji et al. | Chart-text: A fully automated chart image descriptor | |
CN103049763B (en) | Context-constraint-based target identification method | |
CN104408449B (en) | Intelligent mobile terminal scene literal processing method | |
Cheng et al. | Outdoor scene image segmentation based on background recognition and perceptual organization | |
CN108921166A (en) | Medical bill class text detection recognition method and system based on deep neural network | |
CN105574063A (en) | Image retrieval method based on visual saliency | |
CN106529499A (en) | Fourier descriptor and gait energy image fusion feature-based gait identification method | |
CN109523520A (en) | A kind of chromosome automatic counting method based on deep learning | |
CN108268527B (en) | A method of detection land use pattern variation | |
CN105488809A (en) | Indoor scene meaning segmentation method based on RGBD descriptor | |
CN107066972B (en) | Natural scene Method for text detection based on multichannel extremal region | |
Ozdarici-Ok | Automatic detection and delineation of citrus trees from VHR satellite imagery | |
CN104850822B (en) | Leaf identification method under simple background based on multi-feature fusion | |
CN104182763A (en) | Plant type identification system based on flower characteristics | |
CN105069774B (en) | The Target Segmentation method of optimization is cut based on multi-instance learning and figure | |
Tian et al. | Natural scene text detection with MC–MR candidate extraction and coarse-to-fine filtering | |
CN106096542A (en) | Image/video scene recognition method based on range prediction information | |
CN107330027A (en) | A kind of Weakly supervised depth station caption detection method | |
Wang et al. | Remote-sensing image retrieval by combining image visual and semantic features | |
CN106228136A (en) | Panorama streetscape method for secret protection based on converging channels feature | |
CN102609715B (en) | Object type identification method combining plurality of interest point testers | |
CN110458200A (en) | A kind of flower category identification method based on machine learning | |
Huh et al. | Identification of multi-scale corresponding object-set pairs between two polygon datasets with hierarchical co-clustering |
Legal Events
Date | Code | Title | Description |
---|---|---|---|
C06 | Publication | ||
PB01 | Publication | ||
SE01 | Entry into force of request for substantive examination | ||
SE01 | Entry into force of request for substantive examination | ||
REG | Reference to a national code |
Ref country code: HK Ref legal event code: DE Ref document number: 1193181 Country of ref document: HK |
|
GR01 | Patent grant | ||
GR01 | Patent grant | ||
REG | Reference to a national code |
Ref country code: HK Ref legal event code: GR Ref document number: 1193181 Country of ref document: HK |