CN104331898A - Image feature extraction method based on outline sharpness - Google Patents
Image feature extraction method based on outline sharpness Download PDFInfo
- Publication number
- CN104331898A CN104331898A CN201410681339.2A CN201410681339A CN104331898A CN 104331898 A CN104331898 A CN 104331898A CN 201410681339 A CN201410681339 A CN 201410681339A CN 104331898 A CN104331898 A CN 104331898A
- Authority
- CN
- China
- Prior art keywords
- sharpness
- edge
- point
- histogram
- interval
- 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.)
- Pending
Links
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T7/00—Image analysis
- G06T7/10—Segmentation; Edge detection
- G06T7/13—Edge detection
Landscapes
- Engineering & Computer Science (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Physics & Mathematics (AREA)
- General Physics & Mathematics (AREA)
- Theoretical Computer Science (AREA)
- Image Processing (AREA)
- Image Analysis (AREA)
Abstract
The invention discloses an image feature extraction method based on outline sharpness. The method includes: firstly utilizing an edge detection operator to extract the edge of an image; calculating the sharpness of each point on each edge; then working out sharpness histogram vectors of the edges of the image, and at last establishing the overall sharpness histogram of the image. The image features extracted by the method is highly distinguishable and anti-interference. The image feature extraction method can be applied to classifying, identifying and retrieving the images.
Description
Technical field
The present invention relates to a kind of image overall feature extracting method, the aspects such as vehicle mark identification can be applied to.Belong to technical field of image processing.
Background technology
Feature extraction is a concept in computer vision and image procossing.It refers to and uses computing machine to extract image information, determines whether the point of each image belongs to a characteristics of image.The result of feature extraction is that the point on image is divided into different subsets, and these subsets often belong to isolated point, continuous print curve or continuous print region.
Up to the present existing a large amount of image characteristic extracting methods, its feature extracted roughly has following a few class:
(1) color characteristic
The color feature surface nature of image or the scenery corresponding to image-region.General color characteristic is the feature based on pixel, and now all pixels belonging to image or image-region have respective effect.Because the change such as direction, size of Color pair image or image-region is insensitive, so color characteristic can not catch the local feature of objects in images well.
(2) textural characteristics
Textural characteristics describes the surface nature of scenery corresponding to image or image-region.But due to the characteristic that texture is a kind of body surface, the essential attribute of object can not be reflected completely, so only utilize textural characteristics cannot obtain high-level picture material.
(3) spatial relationship feature
Spatial relation description be mutual locus between multiple targets of splitting in image or relative direction relation, these relations can be divided into and comprise/containment relationship, connection/syntople and overlapping/overlapping relation etc.The use of spatial relationship feature can strengthen the description separating capacity to picture material, but spatial relationship feature is often more responsive to the dimensional variation of image or target, rotation and reversion etc.In addition, in practical application, only utilize spatial information can not express scene information effectively and accurately.
Summary of the invention
The object of the invention is to, for existing methodical deficiency, propose a kind of new image characteristic extracting method, to improve the discrimination of characteristics of image.
Technical scheme of the present invention: first utilize edge detection operator to extract the edge of image; Then the sharpness of each point on each edge is calculated; Then obtain the sharpness histogram vectors at each bar edge of image, finally set up the overall sharpness histogram of entire image.
Method of the present invention is realized by following step:
1. extract the edge of image: with edge detection operator, image is processed, obtain the edge of image.
2. the sharpness of each point in edge calculation: the supporting zone first determining each point on edge, then calculates the sharp-pointed angle value of each point.
3. calculate the sharpness histogram vectors at every bar edge: first set the interval number of sharpness distribution and calculate the scope in each interval, then add up the distribution of sharpness in each interval of each point on every bar edge, thus obtain the sharpness histogram vectors at every bar edge.
4. set up overall sharpness histogram: obtain sharpness histogram matrix by the sharpness histogram vectors at each bar edge, by matrix rows phase adduction with belonging to the number of point in each interval divided by the sum that all edges are put, thus obtain final overall sharpness histogram.
Compared with the conventional method, the advantageous of this method exists:
(1) only used the geological information of image edge structure, method is simple and clear, easy to implement;
(2) there is higher counting yield;
(3) compared with traditional histogram feature, the discrimination of sharpness histogram feature is higher;
(4) there is the ability of stronger opposing noise.
(5) can be applicable to the aspects such as the classification of image, identification and retrieval;
For realizing the object of the present invention, be achieved by the following technical solutions:
Based on an image characteristic extracting method for profile sharpness, it is characterized in that comprising following concrete steps:
Step 1: the edge extracting image;
Step 2: the sharpness of each point in edge calculation;
Step 3: the sharpness histogram vectors calculating every bar edge;
Step 4: set up overall sharpness histogram.
Method as above, is characterized in that realizing step 1-4 in the following way:
Step 1: the edge extracting image: with edge detection operator, image is processed, obtain the edge of image;
Step 2: the sharpness of each point in edge calculation: the supporting zone first determining each point on edge, then calculates the sharp-pointed angle value of each point;
Step 3: the sharpness histogram vectors calculating every bar edge: first set the interval number of sharpness distribution and calculate the scope in each interval, then add up the distribution of sharpness in each interval of each point on every bar edge, thus obtain the sharpness histogram vectors at every bar edge;
Step 4: set up overall sharpness histogram: obtain sharpness histogram matrix by the sharpness histogram vectors at each bar edge, by matrix rows phase adduction with belonging to the number of point in each interval divided by the sum that all edges are put, thus obtain final overall sharpness histogram.
Method as above, is characterized in that, described step 2 is the sharp-pointed angle value of each point in edge calculation in the following way: take up an official post in image border and get a bit, use P
irepresent; With a P
icentered by point, get forward and backward k point respectively, (2k+1) that obtain individual point constitutes P
isupporting zone; The span of k is generally 3 ~ 5; α
ifor a P
istrut angle; By two end points P of supporting zone
i-k, P
i+kwith a P
ithe matching that 3 points carry out justifying can obtain the circular arc shown in dotted line, and wherein O point is the center of circle, line segment P
ip
i-kand P
ip
i+kfor sway brace.Because the number of the point of Pi, Pi-k, a Pi+k midfeather is very little, then can suppose | P
ip
i-k|=| P
ip
i+k|, so have:
Work as Pi, when Pi-k, Pi+k are on same straight line, α=180, now have
When α level off to 0 time, have
Define a sharpness variable:
Method as above, is characterized in that, described step 3 calculates the sharpness histogram vectors at every bar edge in the following way:
Suppose that number interval in sharp-pointed histogram is D,
with
represent the upper and lower bound in the interval of m sharpness respectively, they are by formulae discovery below:
For n-th edge S
non every bit, need to judge whether it belongs to interval
then the value of sharpness can just be obtained in interval
interior number, if it is
the number so putting into the point in the interval of m is
suppose that the number that n-th edge is placed into the point in each interval is:
Wherein N
nthe sharp-pointed histogram vectors on n bar limit,
represent the number of the point n-th article of edge belonging to sharply histogrammic m interval, D is interval number total in sharpness histogram.
Method as above, is characterized in that, described step 4 sets up overall sharpness histogram in the following way:
Suppose there is w bar edge in outline map, then sharp-pointed histogram matrix N is expressed as:
Each row of order matrix N is added, and will belong to the number of the point in each interval divided by the sum that all edges are put, thus obtain sharpness histogram vectors F, its expression formula is:
Sharpness histogram is drawn out by sharpness histogram vectors F.
Accompanying drawing explanation
Fig. 1 is the principle schematic that sharpness of the present invention calculates
Fig. 2 is the sharpness histogram of the present invention's " BYD " vehicle mark
Fig. 3 is the sharpness histogram of the present invention's " Ford " vehicle mark
Fig. 4 is the sharpness histogram of the present invention's " Honda " vehicle mark
Fig. 5 is the inventive method schematic flow sheet
Embodiment
In order to understand technical scheme of the present invention better, below in conjunction with drawings and Examples, be described in further detail.
Adopt the inventive method to carry out image characteristics extraction, see accompanying drawing 5, specifically carry out as follows:
1. extract the edge of image
With edge detection operator, image is processed, obtain the edge of image.
2. the sharpness of each point in edge calculation
As shown in Figure 1, accompanying drawing 1 is the enlarged drawing of image border local configuration to its basic skills, and what it was approximate have expressed a P
isupporting zone, α
ifor a P
istrut angle (representing with angle), peripheral solid line is outline line, and stain represents pixel, and dotted line is P
i-k, P
i, P
i+k3 circular arcs fitted to, O point is the center of circle, P
ip
i-k, P
ip
i+kfor sway brace.The value of k generally should be 3-5, some P
i, P
i-k, P
i+kbe approximately 3 points on one section of circular arc, the interval between them is very little, then can suppose | P
ip
i-k|=| P
ip
i+k|, so have
Work as P
i, P
i-k, P
i+ktime on same straight line, α=180, now have
When α level off to 0 time, have
Define a sharpness variable:
Sharp (p
i) represent the acuity of strut angle, sharp (p
i) value is larger, illustrate that this angle is more sharp-pointed, the sharpness sharp (P of every bit on outline line can be calculated according to above formula
i).
3. calculate each bar edge sharpness histogram vectors
The point that have employed sharpness non-zero herein describes global image feature preferably.Suppose that number interval in sharp-pointed histogram is D.
with
represent the upper and lower bound in the interval of m sharpness respectively, they can by formulae discovery below:
For n-th edge S
non every bit, need to judge whether it belongs to interval
then the value of sharpness can just be obtained in interval
interior number, if it is
the number so putting into the point in the interval of m is
suppose that the number that n-th edge is placed into the point in each interval is:
Wherein N
nit is the sharp-pointed histogram vectors on n bar limit.
represent the number of the point n-th article of edge belonging to sharply histogrammic m interval.D is interval number total in sharpness histogram.
4. set up sharpness histogram
Suppose there is w bar edge in outline map, then sharp-pointed histogram matrix N can be expressed as:
Each row of order matrix N is added, and will belong to the number of the point in each interval divided by the sum that all edges are put, thus obtains sharpness histogram vectors F.Its expression formula is:
Global image Feature Descriptor is the row vector of a D dimension.Can draw out sharpness histogram by sharpness histogram vectors F, Fig. 2-4 illustrates the sharpness histogram of various vehicle mark.
Claims (5)
1., based on an image characteristic extracting method for profile sharpness, it is characterized in that comprising following concrete steps:
Step 1: the edge extracting image;
Step 2: the sharpness of each point in edge calculation;
Step 3: the sharpness histogram vectors calculating every bar edge;
Step 4: set up overall sharpness histogram.
2. the method for claim 1, is characterized in that realizing step 1-4 in the following way:
Step 1: the edge extracting image: with edge detection operator, image is processed, obtain the edge of image;
Step 2: the sharpness of each point in edge calculation: the supporting zone first determining each point on edge, then calculates the sharp-pointed angle value of each point;
Step 3: the sharpness histogram vectors calculating every bar edge: first set the interval number of sharpness distribution and calculate the scope in each interval, then add up the distribution of sharpness in each interval of each point on every bar edge, thus obtain the sharpness histogram vectors at every bar edge;
Step 4: set up overall sharpness histogram: obtain sharpness histogram matrix by the sharpness histogram vectors at each bar edge, by matrix rows phase adduction with belonging to the number of point in each interval divided by the sum that all edges are put, thus obtain final overall sharpness histogram.
3. method as claimed in claim 2, is characterized in that, described step 2 is the sharp-pointed angle value of each point in edge calculation in the following way: take up an official post in image border and get a bit, use P
irepresent; With a P
icentered by point, get forward and backward k point respectively, (2k+1) that obtain individual point constitutes P
isupporting zone; The span of k is 3 ~ 5; α
ifor a P
istrut angle; By two end points P of supporting zone
i-k, P
i+kwith a P
ithe matching that 3 points carry out justifying can obtain the circular arc shown in dotted line, and wherein O point is the center of circle, line segment P
ip
i-kand P
ip
i+kfor sway brace, because the number of the point of Pi, Pi-k, a Pi+k midfeather is very little, then can suppose | P
ip
i-k|=| P
ip
i+k|, so have:
Work as Pi, when Pi-k, Pi+k are on same straight line, α=180, now have
When α level off to 0 time, have
Define a sharpness variable:
4. method as claimed in claim 2 or claim 3, it is characterized in that, described step 3 calculates the sharpness histogram vectors at every bar edge in the following way:
Suppose that number interval in sharp-pointed histogram is D,
with
represent the upper and lower bound in the interval of m sharpness respectively, they are by formulae discovery below:
For n-th edge S
non every bit, need to judge whether it belongs to interval
then the value of sharpness is obtained in interval
interior number, if it is
the number so putting into the point in the interval of m is
suppose that the number that n-th edge is placed into the point in each interval is:
Wherein N
nthe sharp-pointed histogram vectors on n bar limit,
represent the number of the point n-th article of edge belonging to sharply histogrammic m interval, D is interval number total in sharpness histogram.
5. the method as described in claim 2 or 4, is characterized in that, described step 4 sets up overall sharpness histogram in the following way:
Suppose there is w bar edge in outline map, then sharp-pointed histogram matrix N is expressed as:
Each row of order matrix N is added, and will belong to the number of the point in each interval divided by the sum that all edges are put, thus obtain sharpness histogram vectors F, its expression formula is:
Sharpness histogram is drawn out by sharpness histogram vectors F.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201410681339.2A CN104331898A (en) | 2014-11-24 | 2014-11-24 | Image feature extraction method based on outline sharpness |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201410681339.2A CN104331898A (en) | 2014-11-24 | 2014-11-24 | Image feature extraction method based on outline sharpness |
Publications (1)
Publication Number | Publication Date |
---|---|
CN104331898A true CN104331898A (en) | 2015-02-04 |
Family
ID=52406618
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN201410681339.2A Pending CN104331898A (en) | 2014-11-24 | 2014-11-24 | Image feature extraction method based on outline sharpness |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN104331898A (en) |
Cited By (1)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN111124113A (en) * | 2019-12-12 | 2020-05-08 | 厦门厦华科技有限公司 | Application starting method based on contour information and electronic whiteboard |
Citations (1)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN101887586A (en) * | 2010-07-30 | 2010-11-17 | 上海交通大学 | Self-adaptive angular-point detection method based on image contour sharpness |
-
2014
- 2014-11-24 CN CN201410681339.2A patent/CN104331898A/en active Pending
Patent Citations (1)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN101887586A (en) * | 2010-07-30 | 2010-11-17 | 上海交通大学 | Self-adaptive angular-point detection method based on image contour sharpness |
Non-Patent Citations (3)
Title |
---|
JIANLI XIAO 等: "Adaptive algorithm for corner detecting based on the degree of sharpness of the contour", 《OPTICAL ENGINEERING》 * |
余旺盛 等: "基于区域边缘统计的图像特征描述新方法", 《计算机学报》 * |
钱文光 等: "基于轮廓尖锐度的图像角点检测算法", 《计算机工程》 * |
Cited By (1)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN111124113A (en) * | 2019-12-12 | 2020-05-08 | 厦门厦华科技有限公司 | Application starting method based on contour information and electronic whiteboard |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
Rong et al. | An improved CANNY edge detection algorithm | |
CN104574375B (en) | Image significance detection method combining color and depth information | |
CN103810503B (en) | Depth study based method for detecting salient regions in natural image | |
CN104835175B (en) | Object detection method in a kind of nuclear environment of view-based access control model attention mechanism | |
CN102722891A (en) | Method for detecting image significance | |
CN106530297A (en) | Object grabbing region positioning method based on point cloud registering | |
CN102521616B (en) | Pedestrian detection method on basis of sparse representation | |
CN104680538B (en) | SAR image CFAR object detection methods based on super-pixel | |
CN104091157A (en) | Pedestrian detection method based on feature fusion | |
CN106682641A (en) | Pedestrian identification method based on image with FHOG- LBPH feature | |
CN104537342B (en) | A kind of express lane line detecting method of combination ridge border detection and Hough transformation | |
CN103577840B (en) | Item identification method | |
CN108109162A (en) | A kind of multiscale target tracking merged using self-adaptive features | |
Bai et al. | Scene text localization using gradient local correlation | |
CN104462382A (en) | Trademark image inquiry method | |
Liang et al. | Robust sea-sky-line detection for complex sea background | |
Fang et al. | ID card identification system based on image recognition | |
CN107133600A (en) | A kind of real-time lane line detection method based on intra-frame trunk | |
CN108090492B (en) | Contour detection method based on scale clue suppression | |
CN103413154A (en) | Human motion identification method based on normalized class Google measurement matrix | |
CN104992448B (en) | The automatic positioning method of robot antisitic defect grape-picking | |
CN104966295B (en) | A kind of naval vessel extracting method based on wire-frame model | |
CN106022342A (en) | Image feature extraction method based on KAZE algorithm | |
CN105354547A (en) | Pedestrian detection method in combination of texture and color features | |
CN108805139A (en) | A kind of image similarity computational methods based on frequency-domain visual significance analysis |
Legal Events
Date | Code | Title | Description |
---|---|---|---|
C06 | Publication | ||
PB01 | Publication | ||
C10 | Entry into substantive examination | ||
SE01 | Entry into force of request for substantive examination | ||
RJ01 | Rejection of invention patent application after publication | ||
RJ01 | Rejection of invention patent application after publication |
Application publication date: 20150204 |