CN102724541B - Intelligent diagnosis and recovery method for monitoring images - Google Patents

Abstract

The invention relates to an intelligent diagnosis and recovery method for monitoring images. The conventional digital monitored image processing system can be only used for diagnosing simple image problems such as brightness, signal deficiency and color offset and can only judge the problems instead of supplying a proper processing sequence and a proper processing method. The intelligent diagnosis and recovery method is characterized by comprising an image problem base, wherein the image problem base lists all possible image degradation problems, and sequences the problems in an inverted sequence direction of the monitored imaging; during running, the problems are sequentially taken out and judged by using the diagnosis method; if a mark of a judgment result is smaller than a preset threshold value, the problem is real, a corresponding processing measure is taken from the list, and the result is sent to a user; if the mark of the judgment result is greater than the set value, the problem is not real, and the next problem is automatically selected and judged; and the steps are repeated until all the problems in the problem base are browsed. By the method, the requirement on professional knowledge of an operator and practice operation for processing the images is reduced; and the operation process is quick and simple.

Description

One kind monitoring image intelligent diagnostics restoration methods

Technical field

The invention belongs to digital image processing field, especially a kind of monitoring image intelligent diagnostics restoration methods. 

Background technology

Digital supervision image can be produced and degraded due to different factors in whole imaging process, and the performance that degrades includes illumination deficiency, colour cast, fuzzy, noise etc., and Questions types are varied and may be overlapped mutually.Sequencing difference is produced because difference degrades（As illumination problem is produced before compression problem）So that the overlay order that difference degrades is also different.When progress degrades the recovery of image, user is firstly the need of all problems for judging monitor video presence；Secondly need to judge the superposition level of each problem according to the process that degrades of video, so that it is determined that recovering the order of processing；Finally suitable processing method is selected to carry out the recovery that degrades for each problem.It can be seen that, before progress degrades recovery operation, it is necessary to first decision problem, determination order, system of selection.That only degradation problems are looked for is accurate, processing sequence is proper, system of selection is suitable, is possible to preferably remove interference, lifts the quality of image. 

Existing digital picture recovers processing system and is often only conceived to processing function, and is not paid close attention to enough for decision problem, determination order, selecting party rule, fully relies on user's experience in itself and Specialized Quality.Which results in three problems：First, it is that product is harsher for the requirement of user, it is necessary to be that the personnel with suitable professional knowledge and operating experience could preferably use；Second, it is so that the process for the recovery that degrades has carried larger subjective colo(u)r, different people, different time, different conditions are likely to be obtained different results；3rd, the propagation of process experience is detrimental to sharing. 

Usual image procossing is to use photoshop softwares, the use of its function needs user to carry out substantial amounts of relational learning and practical operation exercise with reference to help, various operation skills need to be exchanged by third party's mode such as books, network, cause treatment effeciency low, result can not produce a desired effect；Separately there are some digitized video problem diagnosis methods, such as video quality diagnosis system and its implementation（Application number：201110053434.4, applicant：Beijing Vion Technology Co., Ltd.）, the simple image problem such as brightness, signal deletion, colour cast can be only diagnosed, and simply judgement is gone wrong, and suitable processing sequence and method can not be provided. 

The content of the invention

The defect that the present invention exists for prior art, a kind of monitoring image intelligent diagnostics restoration methods are provided, the all problems that monitor video is present can be judged, and rational sorting carries out recovery processing, by unified quality evaluation system and unified process flow, the interference of factor and individual subjective factor in operation can be shielded to a certain extent. 

Therefore, the present invention is adopted the following technical scheme that：One kind monitoring image intelligent diagnostics restoration methods, it is characterised in that including image problem storehouse, all possible image degradation problems are listed in described image problem storehouse, are ranked up each problem with the backward direction of Imaging for Monitoring；During operation, problem is taken out successively, it is judged using diagnostic method, if it is judged that fraction is less than given threshold, then prove that the problem is present, corresponding processing method is now taken out from list, and result is informed into user, if judged result fraction is higher than given threshold, then prove that problem is not present, automatically choose next problem to be judged, circulate above step, until problem is traversed once in problem base. 

The main species of the invention that degrade to monitoring image are summarized, and the presence or absence of every kind of problem and degree are diagnosed respectively by many algorithms, support includes high illumination, low-light (level), contrast deficiency, defocusing blurring, motion blur, random noise, fringes noise, periodic noise, colour cast, blocking effect etc.；The process degraded according to monitoring determines the order recovered, usual Imaging for Monitoring is carried out with the order of target → environment → camera lens → ccd → transmission → compression → storage, its degrade be also according to this sequentially, then carried out when degrading and recovering according to reverse direction, irreversible image being produced to image in view of some recovery operations and primitiveness requirement of the certain methods to image being higher, recovery order is suitably adjusted.Appropriate method is associated for different degradation problems. 

In the presence of a problem is proved, secondary diagnosis is also carried out to the problem and judges that it, with the presence or absence of two grades of problems, is judged two grades of problems using diagnostic method, if judged result fraction is less than given threshold, alignment processing method is then taken out, solution subsequently points to next two grades of problems；If judged result fraction is higher than given threshold, next two grades of problems are pointing directly at；Two grades of problem diagnosis results to the problem judge, if all two grades of problems all solve to finish, are judged automatically into next problem；If not solving all two grades of problems, proceed circulation diagnosis, settlement steps to deal. 

Judge that the fuzziness of image comprises the following steps：1. calculating the Local Extremum of image, the foreground area of image is obtained；2. pair original image carries out the secondary initial value for obscuring, image definition being calculated according to changing rule of the pixel at extreme point under the processing of various fuzzy cores using the Gaussian Blur core of different in width；3. marking a series of Aerial Images of different fog-levels in advance, felt to score one by one according to human eye, a nonlinear fitting is done in definition initial value and human eye vision the effect scoring that step 2 is obtained, the vague marking value after being optimized；4. being finally compared vague marking value with fuzziness threshold value set in advance, that is, obtain fuzzy classification result.

Illumination and contrast problem are diagnosed by foundation of the overall distribution position of image histogram.Carry out illumination when judging when histogram concentrates on low value region for low-light (level), be conversely high illumination；When the more narrow then contrast of histogram range is weaker when contrast judges. 

Colour cast problem is diagnosed with the distributed areas of RGB three channel histograms.During to colour cast Problem judgment, when RGB three channel histograms are when distributing position, scope difference are larger, then it can determine whether as with colour cast. 

For underlying noise, with low pass or the method with passband ripple extract noise contribution；For periodic noise, it is considered to which it detects that the discrete accumulation point of HFS carrys out the stable noise of recognition cycle in spectrogram；For aperiodic fringes noise, separated in the way of contour detecting. 

The problem of described processing method is for high illumination takes enhancing（Over-exposed pattern）, gamma, histogram method processing；For low-light (level) problem, enhancing is taken（Night scene mode）, gamma methods processing；For colour cast problem, white balance, tone homogenizing, the processing of spatial domain colour cast method of adjustment are taken；For defocusing blurring, Gaussian Blur, sharpening method processing are taken；For motion blur, motion blur method is taken to handle；For shake and other fuzzy, blind deblurring, sharpening method processing are taken；For blocking effect, deblocking method is taken to handle；For noise, NLM, mixing spatial domain denoising, the processing of dual-tree complex wavelet denoising method are taken.Be the suitable processing method of various question recommendings with digital picture principle and algorithm principle, such as the method such as the enhancing of illumination question recommending, luminance contrast, histogram；Environmental problem recommends misty rain to remove and night scene Enhancement Method；Colour cast recommends tone homogenizing or white balance method etc..The step can experiential accumulation constantly improve addition. 

The all problems that the present invention exists by all kinds of diagnosis automatic decision monitor videos, according to vision degradation process, provide for the more rational processing sequence of image problem, finally select suitable processing method to carry out the recovery that degrades for each problem.By image problem and corresponding processing method auto-associating, user is recommended, with user-friendly.By the pattern can also from software view Rapid Popularization good process experience and method.Operating process is fast and convenient to be required to the professional knowledge of image procossing and practice operation present invention reduces operator, Problem judgment order determines that method choice is unified specialized, is easy to new technology and the popularization of good experience to share. 

Brief description of the drawings

Fig. 1 is flow chart of the invention

Embodiment

Below by embodiment, technical scheme is described in further detail. 

As shown in figure 1, the handling process of monitoring image intelligent diagnostics restoration methods is as follows, first all possible image degradation problems are listed, each problem are ranked up with the backward direction of Imaging for Monitoring；During operation, problem is taken out successively, it is judged using diagnostic method, if it is judged that fraction is less than given threshold, then prove that the problem is present, corresponding processing method is now taken out from list, and result is informed into user, if judged result fraction is higher than given threshold, then prove that problem is not present, automatically choose next problem to be judged, circulate above step, until problem is traversed once in problem base.In the presence of a problem is proved, secondary diagnosis is also carried out to the problem and judges that it, with the presence or absence of two grades of problems, is judged two grades of problems using diagnostic method, if judged result fraction is less than given threshold, alignment processing method is then taken out, solution subsequently points to next two grades of problems；If judged result fraction is higher than given threshold, next two grades of problems are pointing directly at；Two grades of problem diagnosis results to the problem judge, if all two grades of problems all solve to finish, are judged automatically into next problem；If not solving all two grades of problems, proceed circulation diagnosis, settlement steps to deal. 

The problem of degrading includes high illumination, low-light (level), contrast deficiency, defocusing blurring, motion blur, random noise, fringes noise, periodic noise, colour cast, blocking effect etc.

Wherein judge that the fuzziness of image comprises the following steps：1. calculating the Local Extremum of image, the foreground area of image is obtained；2. pair original image carries out the secondary initial value for obscuring, image definition being calculated according to changing rule of the pixel at extreme point under the processing of various fuzzy cores using the Gaussian Blur core of different in width；3. marking a series of Aerial Images of different fog-levels in advance, felt to score one by one according to human eye, a nonlinear fitting is done in definition initial value and human eye vision the effect scoring that step 2 is obtained, the vague marking value after being optimized；4. being finally compared vague marking value with fuzziness threshold value set in advance, that is, obtain fuzzy classification result. 

Illumination and contrast problem are diagnosed by foundation of the overall distribution position of image histogram.Carry out illumination when judging when histogram concentrates on low value region for low-light (level), be conversely high illumination；When the more narrow then contrast of histogram range is weaker when contrast judges. 

Colour cast problem is diagnosed with the distributed areas of RGB three channel histograms.During to colour cast Problem judgment, when RGB three channel histograms are when distributing position, scope difference are larger, then it can determine whether as with colour cast. 

For underlying noise, with low pass or the method with passband ripple extract noise contribution；For periodic noise, it is considered to which it detects that the discrete accumulation point of HFS carrys out the stable noise of recognition cycle in spectrogram；For aperiodic fringes noise, separated in the way of contour detecting. 

The problem of in restoration methods for high illumination, takes enhancing（Over-exposed pattern）, gamma, histogram method processing；For low-light (level) problem, enhancing is taken（Night scene mode）, gamma methods processing；For colour cast problem, white balance, tone homogenizing, the processing of spatial domain colour cast method of adjustment are taken；For defocusing blurring, Gaussian Blur, sharpening method processing are taken；For motion blur, motion blur method is taken to handle；For shake and other fuzzy, blind deblurring, sharpening method processing are taken；For blocking effect, deblocking method is taken to handle；For noise, NLM, mixing spatial domain denoising, the processing of dual-tree complex wavelet denoising method are taken. 

Of particular note is that, the mode of above-described embodiment is only limitted to describe embodiment, but the present invention is not limited to aforesaid way, and those skilled in the art can easily be modified without departing from the scope of the present invention accordingly, therefore the scope of the present invention should include the maximum magnitude of disclosed principle and new feature. 

Claims (6)

1. one kind monitoring image intelligent diagnostics restoration methods, it is characterised in that including image problem storehouse, all possible image degradation problems are listed in described image problem storehouse, are ranked up each problem with the backward direction of Imaging for Monitoring；During operation, problem is taken out successively, it is judged using diagnostic method, if it is judged that fraction is less than given threshold, then prove that the problem is present, corresponding processing method is now taken out from list, and result is informed into user, if judged result fraction is higher than given threshold, then prove that problem is not present, automatically choose next problem to be judged, circulate above step, until problem is traversed once in problem base；Judge that the fuzziness of image comprises the following steps：A. the Local Extremum of image is calculated, the foreground area of image is obtained；B. the secondary initial value for obscuring, image definition being calculated according to changing rule of the pixel at extreme point under the processing of various fuzzy cores is carried out using the Gaussian Blur core of different in width to original image；C. a series of Aerial Images of different fog-levels are marked in advance, are felt to score one by one according to human eye, the obtained definition initial values of step b and the scoring of human eye vision effect are done into a nonlinear fitting, the vague marking value after being optimized；D. finally vague marking value is compared with fuzziness threshold value set in advance, that is, obtains fuzzy classification result.

2. monitoring image intelligent diagnostics restoration methods according to claim 1, it is characterized in that in the presence of a problem is proved, secondary diagnosis is also carried out to the problem judges it with the presence or absence of two grades of problems, two grades of problems are judged using diagnostic method, if judged result fraction is less than given threshold, alignment processing method is then taken out, solution subsequently points to next two grades of problems；If judged result fraction is higher than given threshold, next two grades of problems are pointing directly at；Two grades of problem diagnosis results to the problem judge, if all two grades of problems all solve to finish, are judged automatically into next problem；If not solving all two grades of problems, proceed circulation diagnosis, settlement steps to deal.

3. monitoring image intelligent diagnostics restoration methods according to claim 2, it is characterised in that diagnosed by foundation of the overall distribution position of image histogram to illumination and contrast problem.

4. monitoring image intelligent diagnostics restoration methods according to claim 3, it is characterised in that diagnosed with the distributed areas of RGB three channel histograms to colour cast problem.

5. it is according to claim 4 monitoring image intelligent diagnostics restoration methods, it is characterised in that for underlying noise, with low pass or the method with passband ripple extract noise contribution；For periodic noise, it is considered to which it detects that the discrete accumulation point of HFS carrys out the stable noise of recognition cycle in spectrogram；For aperiodic fringes noise, separated in the way of contour detecting.

6. monitoring image intelligent diagnostics restoration methods according to claim 5, it is characterised in that the problem of described processing method is for high illumination takes enhancing, gamma, histogram method processing；For low-light (level) problem, enhancing, the processing of gamma methods are taken；For colour cast problem, white balance, tone homogenizing, the processing of spatial domain colour cast method of adjustment are taken；For defocusing blurring, Gaussian Blur, sharpening method processing are taken；For motion blur, motion blur method is taken to handle；For shake and other fuzzy, blind deblurring, sharpening method processing are taken；For blocking effect, deblocking method is taken to handle；For noise, NLM, mixing spatial domain denoising, the processing of dual-tree complex wavelet denoising method are taken.

Images