CN106997583A - A kind of underwater robot adaptive image enhancement and feature extracting method - Google Patents
A kind of underwater robot adaptive image enhancement and feature extracting method Download PDFInfo
- Publication number
- CN106997583A CN106997583A CN201710141451.0A CN201710141451A CN106997583A CN 106997583 A CN106997583 A CN 106997583A CN 201710141451 A CN201710141451 A CN 201710141451A CN 106997583 A CN106997583 A CN 106997583A
- Authority
- CN
- China
- Prior art keywords
- point
- image
- value
- key point
- feature
- 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.)
- Withdrawn
Links
- 238000000034 method Methods 0.000 title claims abstract description 39
- 230000003044 adaptive effect Effects 0.000 title claims abstract description 12
- 238000000605 extraction Methods 0.000 claims abstract description 8
- 238000005516 engineering process Methods 0.000 claims abstract description 7
- 238000010008 shearing Methods 0.000 claims description 12
- 230000002708 enhancing effect Effects 0.000 claims description 6
- 230000004044 response Effects 0.000 claims description 4
- 238000005070 sampling Methods 0.000 claims description 4
- 238000001514 detection method Methods 0.000 claims description 3
- 230000002093 peripheral effect Effects 0.000 claims description 3
- 238000012887 quadratic function Methods 0.000 claims description 3
- 238000012545 processing Methods 0.000 abstract description 7
- 238000011161 development Methods 0.000 abstract description 2
- XLYOFNOQVPJJNP-UHFFFAOYSA-N water Substances O XLYOFNOQVPJJNP-UHFFFAOYSA-N 0.000 description 9
- BBEAQIROQSPTKN-UHFFFAOYSA-N pyrene Chemical compound C1=CC=C2C=CC3=CC=CC4=CC=C1C2=C43 BBEAQIROQSPTKN-UHFFFAOYSA-N 0.000 description 4
- 230000000007 visual effect Effects 0.000 description 4
- 238000005286 illumination Methods 0.000 description 3
- 230000008569 process Effects 0.000 description 3
- 238000003672 processing method Methods 0.000 description 3
- 238000011160 research Methods 0.000 description 3
- 230000003321 amplification Effects 0.000 description 2
- GVEPBJHOBDJJJI-UHFFFAOYSA-N fluoranthrene Natural products C1=CC(C2=CC=CC=C22)=C3C2=CC=CC3=C1 GVEPBJHOBDJJJI-UHFFFAOYSA-N 0.000 description 2
- 239000000463 material Substances 0.000 description 2
- 238000003199 nucleic acid amplification method Methods 0.000 description 2
- 230000003287 optical effect Effects 0.000 description 2
- PXFBZOLANLWPMH-UHFFFAOYSA-N 16-Epiaffinine Natural products C1C(C2=CC=CC=C2N2)=C2C(=O)CC2C(=CC)CN(C)C1C2CO PXFBZOLANLWPMH-UHFFFAOYSA-N 0.000 description 1
- 241000669244 Unaspis euonymi Species 0.000 description 1
- 238000004458 analytical method Methods 0.000 description 1
- 230000009286 beneficial effect Effects 0.000 description 1
- 230000008901 benefit Effects 0.000 description 1
- 230000008859 change Effects 0.000 description 1
- 230000007423 decrease Effects 0.000 description 1
- 230000007123 defense Effects 0.000 description 1
- 238000010586 diagram Methods 0.000 description 1
- 238000003384 imaging method Methods 0.000 description 1
- 239000011159 matrix material Substances 0.000 description 1
- 238000005259 measurement Methods 0.000 description 1
- 238000012544 monitoring process Methods 0.000 description 1
- 230000009466 transformation Effects 0.000 description 1
- 238000013519 translation Methods 0.000 description 1
- 238000012795 verification Methods 0.000 description 1
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T5/00—Image enhancement or restoration
- G06T5/90—Dynamic range modification of images or parts thereof
- G06T5/94—Dynamic range modification of images or parts thereof based on local image properties, e.g. for local contrast enhancement
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T5/00—Image enhancement or restoration
- G06T5/40—Image enhancement or restoration using histogram techniques
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V10/00—Arrangements for image or video recognition or understanding
- G06V10/20—Image preprocessing
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V10/00—Arrangements for image or video recognition or understanding
- G06V10/40—Extraction of image or video features
- G06V10/46—Descriptors for shape, contour or point-related descriptors, e.g. scale invariant feature transform [SIFT] or bags of words [BoW]; Salient regional features
- G06V10/462—Salient features, e.g. scale invariant feature transforms [SIFT]
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V20/00—Scenes; Scene-specific elements
- G06V20/10—Terrestrial scenes
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T2207/00—Indexing scheme for image analysis or image enhancement
- G06T2207/20—Special algorithmic details
- G06T2207/20004—Adaptive image processing
- G06T2207/20012—Locally adaptive
Landscapes
- Engineering & Computer Science (AREA)
- Physics & Mathematics (AREA)
- General Physics & Mathematics (AREA)
- Theoretical Computer Science (AREA)
- Multimedia (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Image Analysis (AREA)
Abstract
The invention discloses a kind of underwater robot adaptive image enhancement in image processing field and feature extracting method, comprise the following steps：1）The video camera carried using underwater robot is carried out image taking and stored；2）Image procossing is carried out using CLAHE technologies；3）Image characteristics extraction is carried out using SIFT algorithms；4）Fixation and recognition is carried out using feature, the underwater identification of underwater robot and positional accuracy are improved by using the present invention, available in marine resources exploratory development.
Description
Technical field
The present invention relates to a kind of underwater robot, more particularly to a kind of underwater robot method for adaptive image enhancement.
Background technology
China is an ocean big country, and the exploration and exploitation of marine resources have weight to the national economy and national defense safety of China
The meaning and value wanted.Energy has been put into machine under water by the research with the mankind to ocean, increasing domestic and foreign scholars
The research of device people.Underwater robot has in fields such as exploration of ocean resources exploitation, marine ecology monitoring, and military affairs extensively should
With.Present underwater robot is provided with greatly optical imagery and image procossing identifying system, and visible images are machines under water
One of primary information resource of device people.But limited by severe imaging circumstances under water, the underwater picture generally existing got
The unfavorable factors such as contrast is low, fuzzy, colour cast.The decline of image visual quality can have a strong impact on subsequent characteristics and extract and target knowledge
The performance of process such as not.Therefore, original underwater picture visual quality is improved by image processing techniques, so as to preferably be schemed
As feature extraction has important scientific meaning, the concern of more and more researchers is obtained, correlative study is increasing.
There are the appearance of some achievements, existing underwater picture processing method to underwater picture processing domestic and foreign scholars in recent years
It is classified largely into image enhaucament and image recovers two classes, because the various methods that image recovers is required for the priori for using environment to know
Know, such as scene depth, locus, optical length, illumination, temperature and electrical conductivity so that during the processing of Image restoration
Between considerably beyond image enchancing method.Therefore image enchancing method has simple and quick, can effectively strengthen the advantage of image.And
In terms of image enhaucament, the histogram equalization (abbreviation HE) based on airspace enhancement is a kind of the most frequently used method, and HE methods are examined
The statistical distribution of each passage value is considered, but has not accounted for positional information, easily strengthened noise and image detail together.Cause
This, in order to overcome the shortcomings of HE methods, is used under water using a kind of contrast limited adaptive histogram equalization CLAHE technologies
Image procossing, its basic thought is the height by limiting local histogram, so as to limit noise amplification and weaken local
Crossing for contrast strengthens.This method can not only stretch the dynamic range of gradation of image, some prominent details, while can pass through
Limit contrast to reduce the distortion phenomenon of image, so as to obtain visual effect better image.
In terms of image characteristics extraction, SIFT (Scale Invariant Feature Transform), i.e. Scale invariant
Eigentransformation, by the continuous exploration of researcher, SIFT description son tools are confirmed after a variety of description are carried out with experimental analysis
There is most strong robustness, can be rotated in image, scale, translate, illumination, carried under the conditions of affine transformation and partial occlusion etc.
Take stable feature.Based on above characteristic, SIFT is obtained for extensively in the fields such as image characteristics extraction, target recognition and tracking
General application, especially for due to underwater complex environment causes shooting image distortion the problem of, can be good at using SIFT algorithms
Solve.Therefore, SIFT algorithms turn into one of domestic and international image procossing and the focus in computer vision research field.
The content of the invention
The purpose of the present invention be for underwater picture due to there are numerous suspended materials in water, light ray energy decay and
The scattering process of pipe, causing image blurring quality to reduce this phenomenon, there is provided a kind of underwater robot adaptive image enhancement
And feature extracting method, to ensure that the view data collected becomes apparent from accurately.
The object of the present invention is achieved like this：A kind of underwater robot adaptive image enhancement and feature extracting method,
Comprise the following steps：
1) video camera carried using underwater robot is carried out image taking and stored；
2) image procossing is carried out using CLAHE technologies；
3) image characteristics extraction is carried out using SIFT algorithms；
4) fixation and recognition is carried out using feature.
Be used as the present invention further restriction, step 2) specific method be：
2-1) by step 1) artwork be divided into k size be m × n continuous nonoverlapping subregion.The value of [m n] is determined
Determine the details enhancing degree of image, generally its value is smaller, and enhancing effect is weaker；
2-2) calculate the grey level histogram per sub-regions；
2-3) determine per sub-regions grey level histogram " shearing " value；
The average value N that number of pixels in per sub-regions is evenly distributed to each gray level is calculated by formula (1) firstaver：
Wherein, NgrayIt is the quantity of gray level in subregion；It is the pixel count in subregion x-axis direction；It is
The pixel count in subregion y-axis direction.Actual " shearing " value is NCL：
NCL=Nclip·Naver (2)
Wherein, NclipIt is interception limit coefficient, it is meant that the pixel count that is included of the limitation each gray level of subregion not
More than the N of mean pixel numberclipTimes；
The grey level histogram per sub-regions 2-4) is sheared, the number of pixels that remainder is come is re-assigned to each histogrammic each
In gray level；
If the sum of all pixels being sheared is NΣclip, then obtain the shearing pixel count N that each gray level is divided equallyacp：
2-5) after redistributing, pixel remainder is shear off for NLP, the step value of pixel is allocated by as follows
Formula is provided：
It is 0 to carry out cycle assignment to pixel count by above-mentioned step value since minimal gray level to residual pixel, final
To new histogram；
2-6) grey level histogram after contrast " shearing " in every sub-regions is equalized；
2-7) according to new histogram distribution, gray scale bilinear interpolation is carried out to image, enhanced result images are generated.
Be used as the present invention further restriction, step 3) specific method be：
3-1) read in step 2-7) in obtained enhanced result images, generate Gaussian difference scale space：For spy
Phenomenon a little unstable on metric space is levied, this problem is solved using Gaussian difference scale space, it is by different scale
Gaussian difference pyrene and image carry out convolution and obtain, such as formula (5)：
D (x, y, σ)=(G (x, y, k σ)-G (x, y, σ)) * I (x, y) (5)
Wherein, σ is that normal distribution standard is poor, and * is convolution symbol, and I (x, y) represents input picture, and G (x, y, σ) is yardstick
Variable Gaussian function, (x, y) represent image pixel position, k be the adjacent yardstick of each two between be separated by linetype scale because
Son；
3-2) metric space extreme point is detected：The acquisition of extreme point is needed by sampled point and certain limit adjacent thereto
Interior point is compared, the scope be under same yardstick around 8 points and totally 26 points of the 2*9 in two neighborhoods up and down, see
Whether the point is maximum or minimum value；
3-3) key point is accurately positioned：Being accurately positioned for key point can improve its matching accuracy, therefore, need to be to key
Point does following processing：
1. position and the dimensional information (reaching sub-pixel) of key point are determined by the fitting of three-dimensional quadratic function；
2. by setting suitable threshold value (typically taking 0.03) to filter out the key point of low contrast；
3. given threshold (generally the taking 10) key point unstable to skirt response is filtered out；
3-4) determine the main gradient direction of key point：According to the scale-value σ for characteristic point and this feature point having been detected by, obtain
To the Gaussian image close to this scale-value：
L (x, y, σ)=G (x, y, σ) * I (x, y) (6)
Gradient magnitude and the direction that the key point can then be obtained are as follows：
M (x, y)=[(L (x+1, y)-L (x-1, y))2+(L(x,y+1)-L(x,y-1))2]1/2 (7)
Wherein, L (x, y, σ) represents the metric space of image, and m (x, y) and θ (x, y) are respectively gaussian pyramid (x, y) place
Gradient magnitude and direction, direction used in L be each crucial point scale under direction；
Obtain after m (x, y) and θ (x, y), sampled in the contiguous range centered on key point, it is counted after sampling
Pixel gradient direction, is then represented with histogram.Histogrammic peak value represents the principal direction of the crucial neighborhood of a point gradient, as
The direction of the key point；So far, feature point detection is finished, and each characteristic point is believed comprising position, yardstick and the aspect of direction 3
Breath；
3-5) generation feature point description：Feature point description is calculated by carrying out image block to key point peripheral region
Histogram of gradients in each piece, generates unique vector descriptor, and specific method is：16 × 16 are taken centered on key point
Sample area, divides the area into 16 seed points that size is 4 × 4, calculates 8 directions respectively in each seed point region
Gradient accumulated value, draw gradient direction direction histogram；The Feature Descriptor of final 128 dimensions for obtaining one 4 × 4 × 8.
Compared with prior art, the beneficial effects of the present invention are：
1) the contrast limited adaptive histogram equalization CLAHE methods employed in this patent are schemed relative to others
As for processing method, by stretching the dynamic range of gradation of image, prominent image detail, while noise amplification is limited, to water
Picture quality can be improved after hypograph processing, and keeps the feature in terms of image detail；
2) the Scale invariant features transform SIFT methods that this patent is used relative to other image characteristic extracting methods come
Say, the feature extracted has translation, rotates and scale the characteristic such as constant, while also having to illumination, noise, small visual angle change
Certain adaptability, in addition, verification and measurement ratio of the characteristic vector extracted under target part circumstance of occlusion is also at a relatively high, it is a kind of
The algorithm of the extraction local feature of robust；
3) image procossing and feature extracting method employed in this patent are under water in the application of robot, under water
Image causes image blurring quality due to there is numerous suspended materials, light ray energy decay and the scattering process of pipe in water
Reducing this phenomenon, there is provided a kind of efficiently feasible method, it is possible to increase the underwater identification of underwater robot and positioning are accurate
Exactness.The present invention can be used in marine resources exploratory development.
Brief description of the drawings
Fig. 1 is workflow diagram of the present invention.
Fig. 2 carries out image processing method flow chart for CLAHE methods in the present invention.
Fig. 3 carries out image characteristic extracting method flow chart for SIFT methods in the present invention.
Embodiment
A kind of underwater robot adaptive image enhancement and feature extracting method as Figure 1-3, comprise the following steps：
1) video camera carried using underwater robot is carried out image taking and stored；
2) image procossing is carried out using CLAHE technologies, specific method is：
2-1) by step 1) artwork be divided into k size be m × n continuous nonoverlapping subregion.The value of [m n] is determined
Determine the details enhancing degree of image, generally its value is smaller, and enhancing effect is weaker；
2-2) calculate the grey level histogram per sub-regions；
2-3) determine per sub-regions grey level histogram " shearing " value；
The average value N that number of pixels in per sub-regions is evenly distributed to each gray level is calculated by formula (1) firstaver：
Wherein, NgrayIt is the quantity of gray level in subregion；It is the pixel count in subregion x-axis direction；It is
The pixel count in subregion y-axis direction.Actual " shearing " value is NCL：
NCL=Nclip·Naver(2)
Wherein, NclipIt is interception limit coefficient, it is meant that the pixel count that is included of the limitation each gray level of subregion not
More than the N of mean pixel numberclipTimes；
The grey level histogram per sub-regions 2-4) is sheared, the number of pixels that remainder is come is re-assigned to each histogrammic each
In gray level；
If the sum of all pixels being sheared is NΣclip, then obtain the shearing pixel count N that each gray level is divided equallyacp：
2-5) after redistributing, pixel remainder is shear off for NLP, the step value of pixel is allocated by as follows
Formula is provided：
It is 0 to carry out cycle assignment to pixel count by above-mentioned step value since minimal gray level to residual pixel, final
To new histogram；
2-6) grey level histogram after contrast " shearing " in every sub-regions is equalized；
2-7) according to new histogram distribution, gray scale bilinear interpolation is carried out to image, enhanced result images are generated；
3) image characteristics extraction is carried out using SIFT algorithms, specific method is：
3-1) read in step 2-7) in obtained enhanced result images, generate Gaussian difference scale space：For spy
Phenomenon a little unstable on metric space is levied, this problem is solved using Gaussian difference scale space, it is by different scale
Gaussian difference pyrene and image carry out convolution and obtain, such as formula (5)：
D (x, y, σ)=(G (x, y, k σ)-G (x, y, σ)) * I (x, y) (5)
Wherein, σ is that normal distribution standard is poor, and * is convolution symbol, and I (x, y) represents input picture, and G (x, y, σ) is yardstick
Variable Gaussian function, (x, y) represent image pixel position, k be the adjacent yardstick of each two between be separated by linetype scale because
Son；
3-2) metric space extreme point is detected：The acquisition of extreme point is needed by sampled point and certain limit adjacent thereto
Interior point is compared, the scope be under same yardstick around 8 points and totally 26 points of the 2*9 in two neighborhoods up and down, see
Whether the point is maximum or minimum value；
3-3) key point is accurately positioned；Key point：Because the extreme point detected is the extreme point of discrete space, and Gauss
Difference operator is more sensitive to border, can produce very strong skirt response, so in the extreme point detected above, it may appear that
Some unstable, low contrast extreme points；In order to remove these pseudo- extreme points, improve and adopted in noise resisting ability, SIFT algorithms
The extreme point of low contrast is got rid of with the method for curve matching, and utilizes the unstable marginal point of Hessian matrix removing.It is surplus
Under point be key point；
Being accurately positioned for key point can improve its matching accuracy, therefore, following processing need to be done to key point：
1. position and the dimensional information (reaching sub-pixel) of key point are determined by the fitting of three-dimensional quadratic function；
2. by setting suitable threshold value (typically taking 0.03) to filter out the key point of low contrast；
3. given threshold (generally the taking 10) key point unstable to skirt response is filtered out；
3-4) determine the main gradient direction of key point：According to the scale-value σ for characteristic point and this feature point having been detected by, obtain
To the Gaussian image close to this scale-value：
L (x, y, σ)=G (x, y, σ) * I (x, y) (6)
Gradient magnitude and the direction that the key point can then be obtained are as follows：
M (x, y)=[(L (x+1, y)-L (x-1, y))2+(L(x,y+1)-L(x,y-1))2]1/2 (7)
Wherein, L (x, y, σ) represents the metric space of image, and m (x, y) and θ (x, y) are respectively gaussian pyramid (x, y) place
Gradient magnitude and direction, direction used in L be each crucial point scale under direction；
Obtain after m (x, y) and θ (x, y), sampled in the contiguous range centered on key point, it is counted after sampling
Pixel gradient direction, is then represented with histogram.Histogrammic peak value represents the principal direction of the crucial neighborhood of a point gradient, as
The direction of the key point；So far, feature point detection is finished, and each characteristic point is believed comprising position, yardstick and the aspect of direction 3
Breath；
3-5) generation feature point description：Feature point description is calculated by carrying out image block to key point peripheral region
Histogram of gradients in each piece, generates unique vector descriptor, and specific method is：16 × 16 are taken centered on key point
Sample area, divides the area into 16 seed points that size is 4 × 4, calculates 8 directions respectively in each seed point region
Gradient accumulated value, draw gradient direction direction histogram；The Feature Descriptor of final 128 dimensions for obtaining one 4 × 4 × 8；
4) fixation and recognition is carried out using feature.
The invention is not limited in above-described embodiment, on the basis of technical scheme disclosed by the invention, the skill of this area
Art personnel are according to disclosed technology contents, it is not necessary to which performing creative labour just can make one to some of which technical characteristic
A little to replace and deform, these are replaced and deformed within the scope of the present invention.
Claims (3)
1. a kind of underwater robot adaptive image enhancement and feature extracting method, it is characterised in that comprise the following steps：
1) video camera carried using underwater robot is carried out image taking and stored；
2) image procossing is carried out using CLAHE technologies；
3) image characteristics extraction is carried out using SIFT algorithms；
4) fixation and recognition is carried out using feature.
2. a kind of underwater robot adaptive image enhancement according to claim 1 and feature extracting method, its feature exist
In step 2) specific method be：
2-1) by step 1) artwork be divided into k size be m × n continuous nonoverlapping subregion.The value of [m n] decides
The details enhancing degree of image, generally its value is smaller, and enhancing effect is weaker；
2-2) calculate the grey level histogram per sub-regions；
2-3) determine per sub-regions grey level histogram " shearing " value；
The average value N that number of pixels in per sub-regions is evenly distributed to each gray level is calculated by formula (1) firstaver：
Wherein, NgrayIt is the quantity of gray level in subregion；It is the pixel count in subregion x-axis direction；It is sub-district
The pixel count in domain y-axis direction.Actual " shearing " value is NCL：
NCL=Nclip·Naver (2)
Wherein, NclipIt is interception limit coefficient, it is flat that it is meant that the pixel count that the limitation each gray level of subregion is included is no more than
The N of equal pixel countclipTimes；
The grey level histogram per sub-regions 2-4) is sheared, the number of pixels that remainder is come is re-assigned to each histogrammic each gray scale
In level；
If the sum of all pixels being sheared is NΣclip, then obtain the shearing pixel count N that each gray level is divided equallyacp：
2-5) after redistributing, pixel remainder is shear off for NLP, the step value of pixel is allocated by equation below
Provide：
It is 0 to carry out cycle assignment to pixel count by above-mentioned step value since minimal gray level to residual pixel, is finally given new
Histogram；
2-6) grey level histogram after contrast " shearing " in every sub-regions is equalized；
2-7) according to new histogram distribution, gray scale bilinear interpolation is carried out to image, enhanced result images are generated.
3. a kind of underwater robot adaptive image enhancement according to claim 2 and feature extracting method, its feature exist
In step 3) specific method be：
3-1) read in step 2-7) in obtained enhanced result images, generate Gaussian difference scale space：For characteristic point
The unstable phenomenon on metric space, this problem is solved using Gaussian difference scale space, it by different scale height
This difference core carries out convolution with image and obtained, such as formula (5)：
D (x, y, σ)=(G (x, y, k σ)-G (x, y, σ)) * I (x, y) (5)
Wherein, σ is that normal distribution standard is poor, and * is convolution symbol, and I (x, y) represents input picture, and G (x, y, σ) is changeable scale
Gaussian function, (x, y) represent image pixel position, k be the adjacent yardstick of each two between be separated by the linetype scale factor；
3-2) metric space extreme point is detected：The acquisition of extreme point is needed by sampled point and adjacent thereto a range of
Point is compared, the scope be under same yardstick around 8 points and totally 26 points of the 2*9 in two neighborhoods up and down, see the point
Whether it is maximum or minimum value；
3-3) key point is accurately positioned：Being accurately positioned for key point can improve its matching accuracy, therefore, need to be done to key point
Handle below：
1. position and the dimensional information of key point are determined by the fitting of three-dimensional quadratic function；
2. by setting suitable threshold value to filter out the key point of low contrast；
3. the given threshold key point unstable to skirt response is filtered out；
3-4) determine the main gradient direction of key point：According to the scale-value σ for characteristic point and this feature point having been detected by, connect
The Gaussian image of this nearly scale-value：
L (x, y, σ)=G (x, y, σ) * I (x, y) (6)
Gradient magnitude and the direction that the key point can then be obtained are as follows：
M (x, y)=[(L (x+1, y)-L (x-1, y))2+(L(x,y+1)-L(x,y-1))2]1/2 (7)
Wherein, L (x, y, σ) represents the metric space of image, and m (x, y) and θ (x, y) are respectively the ladder at gaussian pyramid (x, y) place
Size and Orientation is spent, direction used in L is the direction under each crucial point scale；
Obtain after m (x, y) and θ (x, y), sampled in the contiguous range centered on key point, its pixel is counted after sampling
Gradient direction, is then represented with histogram.Histogrammic peak value represents the principal direction of the crucial neighborhood of a point gradient, is used as the pass
The direction of key point.So far, feature point detection is finished, and each characteristic point includes position, yardstick and the aspect information of direction 3；
3-5) generation feature point description：Feature point description calculates each piece by carrying out image block to key point peripheral region
Interior histogram of gradients, generates unique vector descriptor, and specific method is：Taken centered on key point 16 × 16 sampling
Region, divides the area into 16 seed points that size is 4 × 4, calculates the ladder in 8 directions respectively in each seed point region
Accumulated value is spent, the direction histogram of gradient direction is drawn；The Feature Descriptor of final 128 dimensions for obtaining one 4 × 4 × 8.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201710141451.0A CN106997583A (en) | 2017-03-10 | 2017-03-10 | A kind of underwater robot adaptive image enhancement and feature extracting method |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201710141451.0A CN106997583A (en) | 2017-03-10 | 2017-03-10 | A kind of underwater robot adaptive image enhancement and feature extracting method |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN106997583A true CN106997583A (en) | 2017-08-01 |
Family
ID=59431432
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201710141451.0A Withdrawn CN106997583A (en) | 2017-03-10 | 2017-03-10 | A kind of underwater robot adaptive image enhancement and feature extracting method |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN106997583A (en) |
Cited By (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN111127341A (en) * | 2019-12-05 | 2020-05-08 | Oppo广东移动通信有限公司 | Image processing method and apparatus, and storage medium |
| CN112200019A (en) * | 2020-09-22 | 2021-01-08 | 江苏大学 | Rapid building night scene lighting light fault detection method |
| CN113962890A (en) * | 2021-10-22 | 2022-01-21 | 大连海事大学 | Affine transformation-based underwater color cast image sharpening method |
| CN113989628A (en) * | 2021-10-27 | 2022-01-28 | 哈尔滨工程大学 | Underwater signal lamp positioning method based on weak direction gradient |
Citations (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN103440637A (en) * | 2013-09-22 | 2013-12-11 | 北京智诺英特科技有限公司 | Method and device used for enhancing image sequences |
| CN103839265A (en) * | 2014-02-26 | 2014-06-04 | 西安电子科技大学 | SAR image registration method based on SIFT and normalized mutual information |
| CN104463816A (en) * | 2014-12-02 | 2015-03-25 | 苏州大学 | Image processing method and device |
| CN104794705A (en) * | 2015-03-27 | 2015-07-22 | 努比亚技术有限公司 | Image defogging method and method based on image partial content characteristics |
| CN105809626A (en) * | 2016-03-08 | 2016-07-27 | 长春理工大学 | Self-adaption light compensation video image splicing method |
-
2017
- 2017-03-10 CN CN201710141451.0A patent/CN106997583A/en not_active Withdrawn
Patent Citations (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN103440637A (en) * | 2013-09-22 | 2013-12-11 | 北京智诺英特科技有限公司 | Method and device used for enhancing image sequences |
| CN103839265A (en) * | 2014-02-26 | 2014-06-04 | 西安电子科技大学 | SAR image registration method based on SIFT and normalized mutual information |
| CN104463816A (en) * | 2014-12-02 | 2015-03-25 | 苏州大学 | Image processing method and device |
| CN104794705A (en) * | 2015-03-27 | 2015-07-22 | 努比亚技术有限公司 | Image defogging method and method based on image partial content characteristics |
| CN105809626A (en) * | 2016-03-08 | 2016-07-27 | 长春理工大学 | Self-adaption light compensation video image splicing method |
Non-Patent Citations (1)
| Title |
|---|
| 胡学友: "雾天降质图像的增强复原算法研究", 《中国博士学位论文全文数据库 信息科技辑》 * |
Cited By (6)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN111127341A (en) * | 2019-12-05 | 2020-05-08 | Oppo广东移动通信有限公司 | Image processing method and apparatus, and storage medium |
| CN112200019A (en) * | 2020-09-22 | 2021-01-08 | 江苏大学 | Rapid building night scene lighting light fault detection method |
| CN112200019B (en) * | 2020-09-22 | 2024-02-09 | 上海罗曼照明科技股份有限公司 | Rapid building night scene lighting lamp fault detection method |
| CN113962890A (en) * | 2021-10-22 | 2022-01-21 | 大连海事大学 | Affine transformation-based underwater color cast image sharpening method |
| CN113962890B (en) * | 2021-10-22 | 2024-05-10 | 大连海事大学 | Affine transformation-based underwater color cast image sharpening method |
| CN113989628A (en) * | 2021-10-27 | 2022-01-28 | 哈尔滨工程大学 | Underwater signal lamp positioning method based on weak direction gradient |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN109242888B (en) | Infrared and visible light image fusion method combining image significance and non-subsampled contourlet transformation | |
| CN109948562B (en) | Security check system deep learning sample generation method based on X-ray image | |
| CN109872285B (en) | Retinex low-illumination color image enhancement method based on variational constraint | |
| CN106204509B (en) | Infrared and visible light image fusion method based on regional characteristics | |
| CN106997583A (en) | A kind of underwater robot adaptive image enhancement and feature extracting method | |
| CN110163818A (en) | A kind of low illumination level video image enhancement for maritime affairs unmanned plane | |
| CN106530271B (en) | A kind of infrared image conspicuousness detection method | |
| CN107808161A (en) | A kind of Underwater targets recognition based on light vision | |
| CN108564092A (en) | Sunflower disease recognition method based on SIFT feature extraction algorithm | |
| CN107798670A (en) | A kind of dark primary prior image defogging method using image wave filter | |
| CN109242787A (en) | It paints in a kind of assessment of middle and primary schools' art input method | |
| CN108664970A (en) | A kind of fast target detection method, electronic equipment, storage medium and system | |
| CN110619647A (en) | Method for positioning fuzzy region of image based on combination of edge point frequency domain and spatial domain characteristics | |
| CN108694349A (en) | A kind of pantograph image extraction method and device based on line-scan digital camera | |
| CN113657335A (en) | Mineral phase identification method based on HSV color space | |
| CN105354547A (en) | Pedestrian detection method in combination of texture and color features | |
| CN104992426B (en) | A kind of multi-layer image blending algorithm for light field micro-imaging | |
| CN111626944A (en) | Video deblurring method based on space-time pyramid network and natural prior resistance | |
| CN108960285B (en) | Classification model generation method, tongue image classification method and tongue image classification device | |
| CN111047546A (en) | Infrared image super-resolution reconstruction method and system and electronic equipment | |
| CN110345919A (en) | Space junk detection method based on three-dimensional space vector and two-dimensional plane coordinate | |
| CN114994098A (en) | Foreign matter detection method and device | |
| CN114582017A (en) | Generation method and generation system of gesture data set and storage medium | |
| Cheng et al. | Research on image processing technology based on analog head reading | |
| Tao et al. | Measurement algorithm of notch length of plastic parts based on video |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| WW01 | Invention patent application withdrawn after publication | ||
| WW01 | Invention patent application withdrawn after publication |
Application publication date: 20170801 |