CN102289792A - Method and system for enhancing low-illumination video image - Google Patents
Method and system for enhancing low-illumination video image Download PDFInfo
- Publication number
- CN102289792A CN102289792A CN2011101977496A CN201110197749A CN102289792A CN 102289792 A CN102289792 A CN 102289792A CN 2011101977496 A CN2011101977496 A CN 2011101977496A CN 201110197749 A CN201110197749 A CN 201110197749A CN 102289792 A CN102289792 A CN 102289792A
- Authority
- CN
- China
- Prior art keywords
- image
- low
- retinex
- zone
- level
- 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
Images
Landscapes
- Image Processing (AREA)
Abstract
The invention relates to a method and system for enhancing a low-illumination video image. The method comprises the following steps of: performing brightness delamination on input images according to a Weber-Fechner law to obtain a low-contrast region, a Defrees region, a Weber region and a saturation region; collecting low-illumination high-resolution video images serving as a training set and performing brightness delamination on the images in the training set; adjusting the Retinex parameters of three image sets respectively with an iterative algorithm by taking an image enhancement quality evaluation function as a target function till the value of the target function is optimized; performing image enhancement on a low-contrast region image, a Defrees region image and a Weber region image by using the three obtained parameters; performing image processing on an obtained saturation region image with a gradient field image enhancement algorithm; normalizing the entire gray level range of an image enhancement result pattern; and parallelizing an obtained result and outputting.
Description
Technical field
The present invention relates to image enhancement processing, particularly a kind of low-light (level) video image enhancing method and system.
Background technology
The purpose of figure image intensifying is to improve the visual effect of image, at the application scenario of given image, on purpose emphasizes the integral body or the local characteristics of image, and the difference in the expanded view picture between the different objects feature satisfies the needs of some special analysis.Its method be by certain means to more additional information of original image or transform data, interested feature or inhibition even cover some unwanted feature in the image in the outstanding selectively image are complementary image and eye response characteristic.In figure image intensifying process, do not analyze the reason that picture quality reduces, the image after the processing not necessarily approaches original image.Image enhancement technique is according to the space difference at enhanced processes place, can be divided into based on the algorithm in spatial domain with based on the algorithm two big classes of frequency domain.Directly image gray levels being done computing during based on the algorithm process in spatial domain is in certain transform domain of image the transform coefficient values of image to be carried out certain to revise based on the algorithm of frequency domain, is a kind of algorithm of indirect enhancing.
Strengthen the useful information in the image, it can be the process of a distortion, its objective is to strengthen visual effect.Original unsharp image is become clear or emphasizes some interested feature, suppress uninterested feature, make it to improve picture quality, abundant information amount, strengthen the image processing method of image interpretation and recognition effect.The figure image intensifying can be divided into frequency domain method and space domain method by method therefor.The former regards image as a kind of 2D signal, and it is carried out strengthening based on the signal of two-dimensional Fourier transform.Adopt low pass filtering method, promptly only allow low frequency signal pass through, can remove the noise among the figure; Adopt high-pass filtering method, then can strengthen high frequency signals such as edge, make fuzzy picture become clear.Representative spatial domain algorithm has local averaging method and median filtering method etc., and they can be used for removing or weakening noise.
Algorithm based on the spatial domain is divided into point processing algorithm and neighborhood denoise algorithm.The point processing algorithm is gray level correction, greyscale transformation and histogram modification etc., purpose or make image imaging even, or enlarge dynamic range of images, expanded contrast.The neighborhood enhancement algorithms is divided into two kinds of image smoothing and sharpenings.Level and smooth one is used for the removal of images noise, but also causes the fuzzy of edge easily.Algorithms most in use has mean filter, medium filtering.The purpose of sharpening is the edge contour of outstanding object, is convenient to Target Recognition.Algorithms most in use has gradient method, operator, high-pass filtering, mask matching method, statistics differential technique etc.
Therefore the Enhancement Method versatility of above-mentioned image bottom is stronger, and what bring also is that it does not have specific aim thereupon, down can not be satisfactory as the effect of the figure image intensifying of image irradiation inequality to some environment.With the histogram equalization method is that the greyscale transformation method of representative is a kind of basic skills of handling uneven illumination.The greyscale transformation method refers to uses a certain greyscale transformation function to reach the purpose of compression/stretching gradation of image scope to image.The histogram equalization method is frequency of utilization height in the greyscale transformation method, representative method.It is divided into overall histogram equalization method and partial histogram equalization method.Overall situation histogram equalization method one as the image pre-service, can strengthen effectively.
Picture contrast is remarkable for the low figure image intensifying effect of overall gray value.It is stronger that its details of partial equilibrium method strengthens the more overall histogram equalization method of ability, is applicable to that enhancing need keep the image of original brightness, but problems such as blocking effect, computation rate, enhancing details and enhancing noise can occur.The Retinex algorithm is a kind of algorithm for image enhancement based on illumination compensation, has in contrast the performance that strengthens effect, suppresses excellences such as noise, counting yield.But method be fit to be handled the low image of local gray-value, can effectively strengthen the detail section of dark place wherein and can keep the image original brightness to a certain extent simultaneously in the compressed image contrast.The Retinex Enhancement Method has made full use of the frequency information of image, but does not pay attention to image gradient information and for different images, its model parameter differ greatly difficulty of parameter tuning.The image irradiation inequality is embodied in the Gradient distribution inequality in gradient fields.The picture contrast height shows as big, the clear in structure of gradient intensity, and expanded view can enlarge dynamic range of images as gradient scope, reduces then compressed image dynamic range of gradient scope, therefore can realize the figure image intensifying by handling the image gradient field.The shortcoming of gradient field figure image intensifying is to make image sharpening to a certain extent, and reconstructed image needs certain numerical algorithm in gradient field.
On the application system and product of figure image intensifying, the domestic realtime graphic enhanced system of Beijing Jin Naiwei science and technology limited Company research and development has adopted the Retinex algorithm, and has designed the pattern of misty rain weather.But the advantage of Retinex method is its color constancy, and it changes comparatively significantly regional enhancing effect to brightness in the image can lose a lot of details, also can produce halation phenomenon sometimes.In addition, simple Retinex algorithm is also not enough to the ability of low-light (level) Flame Image Process.Beijing Developmental Technologies LLC of carte blue Creative Science and Technology Co. Ltd has developed the SD Image Intensified System, can handle the ATM of underwater picture and backlight and handle.VIDI
(tall and handsome reaching
TM) will take MotionDSP company by the hand at NVIDIA
(tall and handsome reaching
TM) issue vReveal 2.0 videos of having travelled strengthen software.This software adopts the super-resolution analytical technology, main enhancing towards video image.This system is merely able to move on PC at present in addition.The rich video reinforcement plate that flies the exploitation of electronics technology company limited is integrated histogram transformation, the overall situation and local contrast method of adjustment, these methods all are the basic methods of figure image intensifying.The intelligent video monitoring company of ioimage one tame Israel, video strengthens software and can work under round-the-clock high-noise environments such as sleet mist.
Learn that by above-mentioned investigation and analysis cutting does not also have the real-time enhanced system of HD video on the home market so far, and the method for figure image intensifying is comparatively single, be difficult to satisfy the demand that the complex environment hypograph strengthens.The fundamental research of figure image intensifying also remains further to go deep into and improve.
Summary of the invention
The objective of the invention is to, for addressing the above problem, the present invention proposes a kind of low-light (level) video image enhancing method and system, merging the Retinex algorithm strengthens the property and details and the stereovision of gradient field figure image intensifying in keeping original image to the excellence in low-light (level) zone, and utilize Open MP parallelization treatment technology that view data, algorithm are carried out parallelization and handle, to satisfy the demand that the real-time high-performance of high-definition image strengthens in security protection and monitoring.
For achieving the above object, the present invention proposes a kind of low-light (level) video image enhancing method, it is characterized in that, this method is handled image based on Open MP technology in real time with self-adaptation Retinex algorithm and the combination of gradient field algorithm for image enhancement; The concrete steps of this method comprise:
Step 1): according to the Weber-Fechner law in the psychology, the low-light (level) high-definition image of importing is carried out brightness stratification, obtain low contrast regions, De Vries zone, weber zone and zone of saturation respectively;
Step 2): collect the low-light (level) video image as training set, image in the training set is carried out brightness stratification, image in the training set is divided into low contrast regions, De Vries zone and the weber zone that described step 1) obtains accordingly, obtains three image sets; To scheme image intensifying quality assessment function as objective function, utilize iterative algorithm respectively to three image sets accordingly the Retinex parameter adjust up to the value optimum that reaches objective function;
Step 3): utilize described step 2) three Retinex parameters of Huo Deing adopt the Retinex algorithm to carry out the figure image intensifying to low contrast regions image, De Vries area image and weber area image;
Step 4): the zone of saturation image that described step 1) is obtained adopts the gradient field algorithm for image enhancement to carry out Flame Image Process;
Step 5): respectively the figure as a result after described step 3) and the described step 4) figure image intensifying is passed through respectively the whole normalized of tonal range;
Step 6): the result that described step 5) is obtained carries out exporting after Open MP parallelization is handled; So far, the enhancing in real time of low-light (level) video image is finished.
The method of input low-light (level) high-definition image adopts from the direct input video stream of high definition camera or is stored in video file input the computer in the described step 1).
In the described step 6) image output employing after strengthening is kept at video file in order to the follow-up method of searching and browsing.
For achieving the above object, the present invention also proposes a kind of low-light (level) video image enhanced system, it is characterized in that this system comprises: brightness of image hierarchical block, training Retinex parameter module, Retinex algorithm pattern image intensifying module, gradient field Image Enhancement Based piece, brightness of image is laminated and module and Open MP module;
Described brightness of image hierarchical block is used for the Weber-Fechner law according to psychology, and the low-light (level) high-definition image of importing is carried out brightness stratification, obtains low contrast regions, De Vries zone, weber zone and zone of saturation respectively;
Described training Retinex parameter module, be used to collect the low-light (level) video image as training set, image in the training set is carried out brightness stratification, image in the training set is divided into low contrast regions, De Vries zone and the weber zone that described brightness of image hierarchical block obtains accordingly, obtains three image sets; To scheme image intensifying quality assessment function as objective function, utilize iterative algorithm respectively to three image sets accordingly the Retinex parameter adjust up to the value optimum that reaches objective function;
Described Retinex algorithm pattern image intensifying module, three Retinex parameters that are used to utilize described training Retinex parameter module to obtain adopt the Retinex algorithm to carry out the figure image intensifying to low contrast regions image, De Vries area image and weber area image;
Described gradient field Image Enhancement Based piece, the zone of saturation image that is used for that described brightness of image hierarchical block is obtained adopts the gradient field algorithm for image enhancement to carry out Flame Image Process;
Laminated and the module of described brightness of image is used for respectively the figure as a result after described Retinex algorithm pattern image intensifying module and the described gradient field Image Enhancement Based piece figure image intensifying being passed through respectively the whole normalized of tonal range;
Described Open MP module is used for result described brightness of image is laminated and that module obtains and carries out exporting after Open MP parallelization is handled; So far, the enhancing in real time of low-light (level) video image is finished.
The method of input low-light (level) high-definition image adopts from the direct input video stream of high definition camera or is stored in video file input the computer in the described brightness of image hierarchical block.
In the described Open MP module image output employing after strengthening is kept at video file in order to the follow-up method of searching and browsing.
The invention has the advantages that, adopt Open MP parallelization treatment technology, for strengthening real-time, high clear video image lays a good foundation, the Retinex parameter obtains by continuous iterative learning on the training set widely at one, therefore has certain universality, the artificial parameter of adjusting during image that this algorithm has been avoided gathering under handling different scene images or varying environment has been saved a large amount of labours, has realized the robotization and the intellectuality of system.Carry out area dividing by brightness, be divided into four luminance areas: low contrast regions, De Vries zone, weber zone and zone of saturation high-definition image; Adopt self-adaptation Retinex algorithm to strengthen at first three zone; Image to the zone of saturation carries out the gradient field enhancing; Like this can either the advantage of maximized performance Retinex algorithm aspect low-light (level) figure image intensifying, gradient field can be handled on the performance the best use of again and avoid the halation phenomenon that adopts the Retinex algorithm to cause simultaneously the zone of saturation treatment of picture.
Description of drawings
Fig. 1 is low-light (level) video image enhanced system figure of the present invention;
Fig. 2 is the parallel processing core technology scheme process flow diagram of low-light (level) video image;
Fig. 3 is based on Retinex algorithm and gradient field image technique for enhancing scheme process flow diagram;
Fig. 4 is the high-definition image brightness stratification curve map based on weber-Fechner's law;
Fig. 5 is the image enchancing method process flow diagram based on the Retinex algorithm of image level and the combination of gradient field wild phase;
Fig. 6 is not for carrying out the low-light (level) original image before the figure image intensifying;
Fig. 7 is for carrying out the image after the figure image intensifying with additive method;
Fig. 8 carries out image after the figure image intensifying for the image enchancing method based on the Retinex algorithm of image level and the combination of gradient field wild phase that proposes with the present invention.
Embodiment
Below in conjunction with the drawings and specific embodiments, method of the present invention is described in more detail.
As shown in Figure 1, Fig. 1 is low-light (level) video image enhanced system figure of the present invention.This system has realized that 5,000,000 high clear video images strengthen in real time and are applied in the supervisory system.Realtime graphic enhanced system is particularly based on the enhanced system of high-definition image, and its googol all is strict to the performance demands of enhancement algorithms performance and system according to amount.Enhanced system is to realize on the platform of visual studio, has also adopted the design of hommization in the design of man-machine interaction, has wherein designed the enhancing of various modes, gives the contrast effect that video strengthens front and back.
Except the performance that possesses real-time processing high-definition image, Retinex image enchancing method and gradient field image enchancing method are combined. form the image enchancing method of robust.Input, output and intelligent image that total system mainly comprises the HD video data strengthen three parts.The composition that these three parts are complete the running environment of intelligent video Flame Image Process, set up perfect operation interface in conjunction with visual studio.The user can be convenient, flexible use, exploitation is safeguarded, oneself the algorithm of specific environment hypograph enhancing of upgrade.
Open MP real-time system is based on the real-time high-definition image disposal system of technical grade of windows core and the research and development of visual studio development platform.Because most Image Intensified System is not a real-time high-definition image enhanced system, do not have real-time, can't satisfy has laying and monitoring environment of real-time demand.Treatment technology to mass data is a real-time technical threshold.Appearing as of Open MP technology brought possibility in real time.
As shown in Figure 2, Fig. 2 is the parallel processing core technology scheme process flow diagram of low-light (level) video image.The parallel processing of data and algorithm is carried out in employing to high-definition image based on the parallelization treatment technology of Open MP.The algorithm aspect can realize the multithreading processing on the one hand.Open MP is a kind of multi-threaded parallel treatment technology, is widely used in the large-scale data computing, relates in that image processing field is also rare at present.In the high clear video image widespread use, can bring into play its huge parallel processing capability.By the design of multithreading, can make the operation of the high property of computing machine.The parallel processing of data aspect on the other hand.View data can be regarded a big matrix as, so its data have good concurrency, is fit to very much carry out parallel processing.
At present, many CPU of PC are integrated into Open MP hardware environment are provided, the parallel processing of the Parallel Implementation high-definition image by algorithm, data and thread.
Following table 1 is the test data (Intel (R) coRe (TM) i5750) that the Open MP parallelization of Retinex algorithm for image enhancement is handled:
Table 1 is the test data that the Open MP parallelization of Retinex algorithm for image enhancement is handled
As shown in Figure 3, Fig. 3 is based on Retinex algorithm and gradient field image technique for enhancing scheme process flow diagram.The Retinex algorithm for image enhancement is a kind of algorithm for image enhancement of illumination compensation, its key is the estimation to reflecting component, the whole bag of tricks that proposes is made every effort to strengthen aspects such as effect, inhibition noise, counting yield in contrast and is carried out balance, to reach the best visual effect.This method is fit to handle the low image of local gray-value, can effectively strengthen the wherein detail section of dark place, and can keep the image original brightness to a certain extent simultaneously in the compressed image contrast.Made full use of the frequency information of image based on the Retinex Enhancement Method, but do not paid attention to image gradient information, easily fuzzy edge information when strengthening image.And the image irradiation inequality is embodied in the Gradient distribution inequality in gradient fields.The picture contrast height shows as big, the clear in structure of gradient intensity, and expanded view can enlarge dynamic range of images as gradient scope, reduces then compressed image dynamic range of gradient scope, therefore can realize the figure image intensifying by handling the image gradient field.To sum up, the complementarity of the image enchancing method of Retinex and gradient field image enchancing method is stronger, therefore can effectively be joined together, utilize the Retinex method to come to utilize gradient field to strengthen the marginal information at the bright place of image at the detail section that strengthens the dark place.
As shown in Figure 4, Fig. 4 is the high-definition image brightness stratification curve map based on weber-Fechner's law.Image to be strengthened one to present contrast low, the too dark brightness of topography, also or topography's brightness strong excessively.According to the weber-Fechner's law Δ I/I=K in the psychology, I represents the intensity that stimulates, the just noticeable difference that Δ I representative stimulates, K is a constant, also being called weber fraction. slope of a curve is 1/2 in figure below, just weber fraction is that 1/2 zone is called the De Vries zone, is the low-light (level) zone; Slope is that 1 zone is famous weber zone, is middle illumination zone; And because Weber's law only is fit to stimulus intensity is not under the very big situation, the saturated influence of high illumination zone irriate is defined as the zone of saturation; Sheng Xia zone is exactly a low contrast regions at last, the almost variation of imperceptible illumination of human eye in this zone.
As shown in Figure 5, Fig. 5 is based on the Retinex algorithm of image level and the image enchancing method process flow diagram of gradient field wild phase combination.The joint of Retinex algorithm and gradient field enhancement algorithms is the performance advantage separately of degree greatly, learns from other's strong points to offset one's weaknesses.
Though the Retinex algorithm has superior image to strengthen the property, its parameter needs to adjust in the different images.Come an also non-gravy jobs for the technical professional, say nothing of the layman.Widespread use for ease of performance Retinex algorithm has proposed self-adaptation Retinex parameter estimation and enhancement algorithms based on pattern-recognition.Self-adaptation Retinex algorithm is adopted in low contrast regions, De Vries zone and weber zone.Utilize the multitude of video image of collecting as training set, the image in the training set is carried out layering and be divided into three above-mentioned area images concentrating.Utilize figure image intensifying quality assessment function as objective function, utilize iterative algorithm that the value of objective function is adjusted and detected to the Retinex parameter, when reaching objective function optimum, stop iteration and export the parameter of Retinex.The parameter that obtains obtains by continuous iterative learning on the training set widely at one, therefore has certain universality.
Self-adaptation Retinex algorithm becomes different parameter points with parameter quantification, and each point is all corresponding to one group of different in parameter space parameters, and every group of different parameter all is used in the figure image intensifying under variety classes and the environment.At first the image on the high-definition image training set is carried out the rough segmentation piece and carry out cluster, obtain concentrating the image training set under the varying environment.Obtain its optimal parameter by utilizing the Retinex algorithm to carry out iteration at each training set.This parameter is to arrive in the acquistion of going to school of a large amount of training sets, has certain universality and robustness.Algorithm in the past all is to carry out careful parameter adjustment could obtain better to strengthen effect on single image.The artificial parameter of adjusting during image that this algorithm has been avoided gathering under handling different scene images or varying environment has been saved a large amount of labours, has realized the robotization and the intellectuality of system.
Another implication of self-adaptation Retinex is to be good at the different parameter processing targetedly respectively of image-region employing of processing at the Retinex algorithm.Adopt three kinds of different parameters respectively at contrasted zones, De Vries zone and weber zone, these parameters all are to obtain by above-mentioned study.
The method of Retinex theory can be compressed the dynamic range of uneven illumination image effectively, increases in real time in the image than the contrast of dark place, improves the visual effect of image effectively; But, shortcoming of these method ubiquities, promptly when improving the integral image effect, strengthening than the dark place details, the details at brighter place often can not get strengthening in the original image, even weakened.So adopt the gradient field enhancement techniques that the zone of saturation is strengthened separately, this method combines the human visual perception characteristic contrast at brighter place in the image is carried out bigger enhancing, has solved the problems referred to above effectively.
Suppress bigger gradient,, reduce the influence of uneven illumination with the dynamic range of compressed image; The less gradient that stretches is to strengthen image detail.This model can be expressed as:
G(x，y)＝ΔH(x，y)Ψ(x，y) (1)
Wherein, (x y) is the gradient of original image to Δ H, and (x is that it depends on the size of original image partial gradient to the function of gradient fields operation y), and (x y) is the image gradient field after strengthening to G.After the gradient fields that strengthens, can obtain the image reconstruction under the least square meaning.
∫∫F(ΔI，G)dxdy (2)
Wherein,
Proposed the algorithm for image enhancement that a kind of Retinex algorithm combines with the gradient field wild phase, wherein the Retinex algorithm is adaptive Retinex algorithm.Carry out area dividing by brightness, be divided into four luminance areas: low contrast regions, De Vries zone, weber zone and zone of saturation high-definition image; Adopt self-adaptation Retinex algorithm to strengthen at first three zone; Image to the zone of saturation carries out the gradient field enhancing; Like this can either the advantage of maximized performance Retinex algorithm aspect low-light (level) figure image intensifying, gradient field can be handled on the performance the best use of again and avoid the halation phenomenon that adopts the Retinex algorithm to cause simultaneously the zone of saturation treatment of picture.Traditional method is that image is carried out whole processing is irrational, and particularly for high-definition image, the detailed information of image local, exposure status etc. are uneven, therefore also just cannot treat different things as the same, and need divide and rule.
The tonal range of image may be different with former figure after the figure image intensifying of zones of different, and may occur exceeding 255, by whole normalization is adjusted into identically with former figure to tonal range, and realize that image physically merges.
Adopt two kinds of inputs in this system, a kind of is the video flowing of directly importing from the high definition camera; A kind of is the video file that is stored in the computer; Dual mode satisfies different sight demands.At output facet, the image after the enhancing can be kept at video file in order to follow-up searching and browsing.
The real-time enhanced system of intelligence high clear video image is innovated on traditional images Enhancement Method basis and is improved.Proposed the high-definition image enhancing first and built a real-time enhanced system from system aspects.One is the problem that solves Retinex parameter adjustment difficulty aspect algorithm; Set up model image division is become different luminance levels, adopt different algorithms comprising self-adaptation Retinex algorithm and gradient field enhancement algorithms at the image of different levels.Aspect the enhancing of quickening the realization high-definition image, adopt Open MP concurrent technique that enhancement algorithms is quickened to handle.
As shown in Figure 6, Fig. 6 is not for carrying out the low-light (level) original image before the figure image intensifying.As shown in Figure 7, Fig. 7 is for carrying out the image after the figure image intensifying with additive method.As shown in Figure 8, Fig. 8 carries out image after the figure image intensifying for the image enchancing method based on the Retinex algorithm of image level and the combination of gradient field wild phase that proposes with the present invention.Contrast as can be known from Fig. 6, Fig. 7 and Fig. 8, can either the advantage of maximized performance Retinex algorithm aspect low-light (level) figure image intensifying, gradient field can be handled on the performance the best use of again and avoid the halation phenomenon that adopts the Retinex algorithm to cause simultaneously the zone of saturation treatment of picture.
It should be noted last that above embodiment is only unrestricted in order to technical scheme of the present invention to be described.Although the present invention is had been described in detail with reference to embodiment, those of ordinary skill in the art is to be understood that, technical scheme of the present invention is made amendment or is equal to replacement, do not break away from the spirit and scope of technical solution of the present invention, it all should be encompassed in the middle of the claim scope of the present invention.
Claims (6)
1. a low-light (level) video image enhancing method is characterized in that, this method is handled image based on Open MP technology in real time with self-adaptation Retinex algorithm and the combination of gradient field algorithm for image enhancement; The concrete steps of this method comprise:
Step 1): according to the Weber-Fechner law in the psychology, the low-light (level) high-definition image of importing is carried out brightness stratification, obtain low contrast regions, De Vries zone, weber zone and zone of saturation respectively;
Step 2): collect the low-light (level) video image as training set, image in the training set is carried out brightness stratification, image in the training set is divided into low contrast regions, De Vries zone and the weber zone that described step 1) obtains accordingly, obtains three image sets; To scheme image intensifying quality assessment function as objective function, utilize iterative algorithm respectively to three image sets accordingly the Retinex parameter adjust up to the value optimum that reaches objective function;
Step 3): utilize described step 2) three Retinex parameters of Huo Deing adopt the Retinex algorithm to carry out the figure image intensifying to low contrast regions image, De Vries area image and weber area image;
Step 4): the zone of saturation image that described step 1) is obtained adopts the gradient field algorithm for image enhancement to carry out Flame Image Process;
Step 5): respectively the figure as a result after described step 3) and the described step 4) figure image intensifying is passed through respectively the whole normalized of tonal range;
Step 6): the result that described step 5) is obtained carries out exporting after Open MP parallelization is handled; So far, the enhancing in real time of low-light (level) video image is finished.
2. low-light (level) video image enhancing method according to claim 1 is characterized in that, the method for input low-light (level) high-definition image adopts from the direct input video stream of high definition camera or is stored in video file input the computer in the described step 1).
3. low-light (level) video image enhancing method according to claim 1 is characterized in that, in the described step 6) image output employing after strengthening is kept at video file in order to the follow-up method of searching and browsing.
4. low-light (level) video image enhanced system, it is characterized in that this system comprises: brightness of image hierarchical block, training Retinex parameter module, Retinex algorithm pattern image intensifying module, gradient field Image Enhancement Based piece, brightness of image is laminated and module and Open MP module;
Described brightness of image hierarchical block is used for the Weber-Fechner law according to psychology, and the low-light (level) high-definition image of importing is carried out brightness stratification, obtains low contrast regions, De Vries zone, weber zone and zone of saturation respectively;
Described training Retinex parameter module, be used to collect the low-light (level) video image as training set, image in the training set is carried out brightness stratification, image in the training set is divided into low contrast regions, De Vries zone and the weber zone that described brightness of image hierarchical block obtains accordingly, obtains three image sets; To scheme image intensifying quality assessment function as objective function, utilize iterative algorithm respectively to three image sets accordingly the Retinex parameter adjust up to the value optimum that reaches objective function;
Described Retinex algorithm pattern image intensifying module, three Retinex parameters that are used to utilize described training Retinex parameter module to obtain adopt the Retinex algorithm to carry out the figure image intensifying to low contrast regions image, De Vries area image and weber area image;
Described gradient field Image Enhancement Based piece, the zone of saturation image that is used for that described brightness of image hierarchical block is obtained adopts the gradient field algorithm for image enhancement to carry out Flame Image Process;
Laminated and the module of described brightness of image is used for respectively the figure as a result after described Retinex algorithm pattern image intensifying module and the described gradient field Image Enhancement Based piece figure image intensifying being passed through respectively the whole normalized of tonal range;
Described Open MP module is used for result described brightness of image is laminated and that module obtains and carries out exporting after Open MP parallelization is handled; So far, the enhancing in real time of low-light (level) video image is finished.
5. low-light (level) video image enhanced system according to claim 4, it is characterized in that the method for input low-light (level) high-definition image adopts from the direct input video stream of high definition camera or is stored in video file input the computer in the described brightness of image hierarchical block.
6. low-light (level) video image enhanced system according to claim 4 is characterized in that, in the described Open MP module image output employing after strengthening is kept at video file in order to the follow-up method of searching and browsing.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN2011101977496A CN102289792A (en) | 2011-05-03 | 2011-07-14 | Method and system for enhancing low-illumination video image |
Applications Claiming Priority (3)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201110113205.7 | 2011-05-03 | ||
| CN201110113205 | 2011-05-03 | ||
| CN2011101977496A CN102289792A (en) | 2011-05-03 | 2011-07-14 | Method and system for enhancing low-illumination video image |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN102289792A true CN102289792A (en) | 2011-12-21 |
Family
ID=45336184
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN2011101977496A Pending CN102289792A (en) | 2011-05-03 | 2011-07-14 | Method and system for enhancing low-illumination video image |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN102289792A (en) |
Cited By (23)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN103020924A (en) * | 2012-12-13 | 2013-04-03 | 武汉大学 | Low-illumination monitored image enhancement method based on similar scenes |
| CN103037208A (en) * | 2012-12-20 | 2013-04-10 | 上海市电力公司 | Recovering method of digital video of underground substation |
| CN103337056A (en) * | 2013-07-05 | 2013-10-02 | 中国民用航空总局第二研究所 | Adaptive illumination compressed image intensification algorithm |
| CN103386835A (en) * | 2012-05-11 | 2013-11-13 | 索尼电脑娱乐欧洲有限公司 | Augmented reality system |
| CN104346776A (en) * | 2013-08-02 | 2015-02-11 | 杭州海康威视数字技术股份有限公司 | Retinex-theory-based nonlinear image enhancement method and system |
| CN104346609A (en) * | 2013-08-01 | 2015-02-11 | 阿里巴巴集团控股有限公司 | Method and device for recognizing characters on printed products |
| CN104392436A (en) * | 2014-11-11 | 2015-03-04 | 莱芜钢铁集团有限公司 | Processing method and device for remote sensing image |
| CN104700111A (en) * | 2013-12-04 | 2015-06-10 | 华平信息技术股份有限公司 | Method and system for vehicle color identification based on Retinex image enhancement algorithm |
| CN104809700A (en) * | 2015-04-16 | 2015-07-29 | 北京工业大学 | Low-light video real-time enhancement method based on bright channel |
| CN104809712A (en) * | 2015-05-15 | 2015-07-29 | 河海大学常州校区 | Rapid image repairing method based on rough set |
| CN104917969A (en) * | 2015-05-30 | 2015-09-16 | 广东欧珀移动通信有限公司 | Image processing method and mobile terminal |
| CN106875383A (en) * | 2017-01-24 | 2017-06-20 | 北京理工大学 | The insensitive blurred picture quality evaluating method of content based on Weibull statistical nature |
| CN107832967A (en) * | 2017-11-23 | 2018-03-23 | 福建农林大学 | A kind of sound scape degrees of coordination dynamic evaluation method suitable for bamboo grove space |
| CN108122210A (en) * | 2017-12-19 | 2018-06-05 | 长沙全度影像科技有限公司 | A kind of low-light (level) car plate video image enhancing method based on Retinex and enhancing gradient |
| CN109035228A (en) * | 2018-07-18 | 2018-12-18 | 中北大学 | A kind of radioscopic image processing method of non-uniform thickness component |
| CN109785240A (en) * | 2017-11-13 | 2019-05-21 | ***通信有限公司研究院 | A kind of enhancement method of low-illumination image, device and image processing equipment |
| CN109785239A (en) * | 2017-11-13 | 2019-05-21 | 华为技术有限公司 | The method and apparatus of image procossing |
| CN111476725A (en) * | 2020-03-24 | 2020-07-31 | 广西科技大学 | Image defogging enhancement algorithm based on gradient domain oriented filtering and multi-scale Retinex theory |
| CN111784582A (en) * | 2020-07-08 | 2020-10-16 | 桂林电子科技大学 | DEC-SE-based low-illumination image super-resolution reconstruction method |
| CN112116807A (en) * | 2020-08-31 | 2020-12-22 | 广西交科集团有限公司 | Multifunctional traffic safety guiding device |
| CN112565718A (en) * | 2020-12-01 | 2021-03-26 | 大连海洋大学 | Underwater image processing method based on Retinex and high dynamic range image gradient compression |
| CN113284061A (en) * | 2021-05-17 | 2021-08-20 | 大连海事大学 | Underwater image enhancement method based on gradient network |
| CN115100077A (en) * | 2022-07-25 | 2022-09-23 | 深圳市安科讯实业有限公司 | Novel image enhancement method and device |
-
2011
- 2011-07-14 CN CN2011101977496A patent/CN102289792A/en active Pending
Non-Patent Citations (4)
| Title |
|---|
| 刘昕等: "多核并行空间域图像增强算法", 《微计算机信息》 * |
| 李福文等: "基于Ｒｅｔｉｎｅｘ模型的彩色图像全局增强算法", 《北京理工大学学报》 * |
| 梁琳等: "光照不均图像增强方法综述", 《计算机应用研究》 * |
| 汪荣贵等: "基于照度分割的局部多尺度Retinex算法", 《电子学报》 * |
Cited By (38)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN103386835A (en) * | 2012-05-11 | 2013-11-13 | 索尼电脑娱乐欧洲有限公司 | Augmented reality system |
| CN103020924A (en) * | 2012-12-13 | 2013-04-03 | 武汉大学 | Low-illumination monitored image enhancement method based on similar scenes |
| CN103020924B (en) * | 2012-12-13 | 2015-03-25 | 武汉大学 | Low-illumination monitored image enhancement method based on similar scenes |
| CN103037208B (en) * | 2012-12-20 | 2015-09-09 | 上海市电力公司 | A kind of digital video of underground substation restoration methods |
| CN103037208A (en) * | 2012-12-20 | 2013-04-10 | 上海市电力公司 | Recovering method of digital video of underground substation |
| CN103337056A (en) * | 2013-07-05 | 2013-10-02 | 中国民用航空总局第二研究所 | Adaptive illumination compressed image intensification algorithm |
| CN103337056B (en) * | 2013-07-05 | 2015-12-09 | 中国民用航空总局第二研究所 | A kind of adaptive optical is according to compressed image Enhancement Method |
| CN104346609A (en) * | 2013-08-01 | 2015-02-11 | 阿里巴巴集团控股有限公司 | Method and device for recognizing characters on printed products |
| CN104346609B (en) * | 2013-08-01 | 2018-05-04 | 阿里巴巴集团控股有限公司 | The method and device of character on a kind of identification printed matter |
| CN104346776A (en) * | 2013-08-02 | 2015-02-11 | 杭州海康威视数字技术股份有限公司 | Retinex-theory-based nonlinear image enhancement method and system |
| CN104346776B (en) * | 2013-08-02 | 2017-05-24 | 杭州海康威视数字技术股份有限公司 | Retinex-theory-based nonlinear image enhancement method and system |
| CN104700111A (en) * | 2013-12-04 | 2015-06-10 | 华平信息技术股份有限公司 | Method and system for vehicle color identification based on Retinex image enhancement algorithm |
| CN104392436A (en) * | 2014-11-11 | 2015-03-04 | 莱芜钢铁集团有限公司 | Processing method and device for remote sensing image |
| CN104809700A (en) * | 2015-04-16 | 2015-07-29 | 北京工业大学 | Low-light video real-time enhancement method based on bright channel |
| CN104809700B (en) * | 2015-04-16 | 2017-08-25 | 北京工业大学 | A kind of low-light (level) video real time enhancing method based on bright passage |
| CN104809712A (en) * | 2015-05-15 | 2015-07-29 | 河海大学常州校区 | Rapid image repairing method based on rough set |
| CN104809712B (en) * | 2015-05-15 | 2018-01-16 | 河海大学常州校区 | A kind of image fast repairing method based on rough set |
| CN104917969B (en) * | 2015-05-30 | 2018-01-19 | 广东欧珀移动通信有限公司 | The method and mobile terminal of a kind of image procossing |
| CN104917969A (en) * | 2015-05-30 | 2015-09-16 | 广东欧珀移动通信有限公司 | Image processing method and mobile terminal |
| CN106875383B (en) * | 2017-01-24 | 2020-05-08 | 北京理工大学 | Content insensitive fuzzy image quality evaluation method based on Weibull statistical characteristics |
| CN106875383A (en) * | 2017-01-24 | 2017-06-20 | 北京理工大学 | The insensitive blurred picture quality evaluating method of content based on Weibull statistical nature |
| CN109785240B (en) * | 2017-11-13 | 2021-05-25 | ***通信有限公司研究院 | Low-illumination image enhancement method and device and image processing equipment |
| CN109785239B (en) * | 2017-11-13 | 2021-05-04 | 华为技术有限公司 | Image processing method and device |
| CN109785240A (en) * | 2017-11-13 | 2019-05-21 | ***通信有限公司研究院 | A kind of enhancement method of low-illumination image, device and image processing equipment |
| CN109785239A (en) * | 2017-11-13 | 2019-05-21 | 华为技术有限公司 | The method and apparatus of image procossing |
| CN107832967B (en) * | 2017-11-23 | 2021-09-14 | 福建农林大学 | Sound scene coordination degree dynamic evaluation method suitable for bamboo forest space |
| CN107832967A (en) * | 2017-11-23 | 2018-03-23 | 福建农林大学 | A kind of sound scape degrees of coordination dynamic evaluation method suitable for bamboo grove space |
| CN108122210A (en) * | 2017-12-19 | 2018-06-05 | 长沙全度影像科技有限公司 | A kind of low-light (level) car plate video image enhancing method based on Retinex and enhancing gradient |
| CN109035228A (en) * | 2018-07-18 | 2018-12-18 | 中北大学 | A kind of radioscopic image processing method of non-uniform thickness component |
| CN109035228B (en) * | 2018-07-18 | 2021-12-28 | 中北大学 | X-ray image processing method of non-uniform-thickness component |
| CN111476725A (en) * | 2020-03-24 | 2020-07-31 | 广西科技大学 | Image defogging enhancement algorithm based on gradient domain oriented filtering and multi-scale Retinex theory |
| CN111784582A (en) * | 2020-07-08 | 2020-10-16 | 桂林电子科技大学 | DEC-SE-based low-illumination image super-resolution reconstruction method |
| CN111784582B (en) * | 2020-07-08 | 2022-09-27 | 桂林电子科技大学 | DEC-SE-based low-illumination image super-resolution reconstruction method |
| CN112116807A (en) * | 2020-08-31 | 2020-12-22 | 广西交科集团有限公司 | Multifunctional traffic safety guiding device |
| CN112565718A (en) * | 2020-12-01 | 2021-03-26 | 大连海洋大学 | Underwater image processing method based on Retinex and high dynamic range image gradient compression |
| CN113284061A (en) * | 2021-05-17 | 2021-08-20 | 大连海事大学 | Underwater image enhancement method based on gradient network |
| CN113284061B (en) * | 2021-05-17 | 2024-04-05 | 大连海事大学 | Underwater image enhancement method based on gradient network |
| CN115100077A (en) * | 2022-07-25 | 2022-09-23 | 深圳市安科讯实业有限公司 | Novel image enhancement method and device |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN102289792A (en) | Method and system for enhancing low-illumination video image | |
| CN101452575B (en) | Image self-adapting enhancement method based on neural net | |
| Ju et al. | Single image dehazing via an improved atmospheric scattering model | |
| US10152781B2 (en) | Method for image processing using local statistics convolution | |
| Zhu et al. | An adaptive histogram equalization algorithm on the image gray level mapping | |
| CN101833754B (en) | Image enhancement method and image enhancement system | |
| CN108765336B (en) | Image defogging method based on dark and bright primary color prior and adaptive parameter optimization | |
| CN104346776B (en) | Retinex-theory-based nonlinear image enhancement method and system | |
| CN100478993C (en) | Image reinforcement method for self-adaptive regulation according to edge and brightness | |
| CN102231791B (en) | Video image defogging method based on image brightness stratification | |
| CN102096909B (en) | Improved unsharp masking image reinforcing method based on logarithm image processing model | |
| CN102625030B (en) | video enhancement method and system | |
| CN108122206A (en) | A kind of low-light (level) image denoising method and device | |
| CN105046677A (en) | Enhancement processing method and apparatus for traffic video image | |
| Zhao et al. | Single image fog removal based on local extrema | |
| CN104182947A (en) | Low-illumination image enhancement method and system | |
| CN101853498A (en) | Image synthetizing method and image processing device | |
| CN103295204B (en) | A kind of image self-adapting enhancement method based on non-down sampling contourlet transform | |
| CN105427257A (en) | Image enhancement method and apparatus | |
| CN105046658A (en) | Low-illumination image processing method and device | |
| CN102750682A (en) | Image preprocessing method for processing nonuniform illumination of miner face image and coal surface | |
| CN102063712A (en) | Multi-exposure image fusion method based on sub-band structure | |
| CN103679157A (en) | Human face image illumination processing method based on retina model | |
| CN104424641A (en) | Detection method for image fuzzy tampering | |
| CN103489168A (en) | Enhancing method and system for infrared image being converted to pseudo color image in self-adaptive mode |
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 | ||
| RJ01 | Rejection of invention patent application after publication | ||
| RJ01 | Rejection of invention patent application after publication |
Application publication date: 20111221 |