Image correction optimization method, system, device and storage medium

Abstract

The application discloses an image correction optimization method, system, device and storage medium, and technical scheme points thereof are as follows: determining a first distortion parameter and a first image attribute of a current text image in an image acquisition device; screening secondary processing records from historical image processing records according to the first image attribute; generating a first correction coefficient according to a correction law that a historical correction value of automatic white balance in all secondary processing records changes over time; determining an actual correction value of automatic white balance of the current text image, determining a predicted correction value of the current text image according to the correction law, and generating a second correction coefficient according to a deviation between the actual correction value and the predicted correction value; and correcting the first distortion parameter according to the first correction coefficient and the second correction coefficient to obtain a second distortion parameter. According to the application, the actual device state and the instant deviation can be accurately reflected through the automatic white balance correction value, and the distortion parameter is adjusted, so that the accuracy and robustness of image preprocessing are improved.

Description

TECHNICAL FIELD

[0001] The application belongs to the technical field of image processing, and particularly relates to an image correction optimization method, system, device and storage medium. BACKGROUND

[0002] At present, image correction technology is widely used in fields such as document digitization, automatic monitoring and intelligent transportation, especially in machine vision systems, image preprocessing is an important link to ensure the accuracy of subsequent character recognition. The existing technology mainly relies on fixed algorithms or simple adaptive methods to correct the geometric distortion in the image, but these methods often process based on single image data, and cannot fully capture the influence of performance degradation of the acquisition device and environmental changes on image quality.

[0003] However, since the image analysis device and the image acquisition device usually belong to different systems, the image analysis device cannot directly obtain the state information of the image acquisition device, which leads to the fact that the existing image correction method cannot dynamically compensate for local distortion caused by device performance decline in the preprocessing stage. In the existing technology, the application of automatic white balance correction mainly adopts static parameter setting or simple real-time adjustment, and cannot analyze and predict the long-term performance changes of the device using the time series data in the historical image processing records, so as to accurately reflect the actual device state and instantaneous deviation in the processing process. SUMMARY

[0004] The purpose of the present application is to provide an image correction optimization method, system, device and storage medium, which accurately reflects the actual device state and instantaneous deviation through automatic white balance correction value, and adjusts the distortion parameter accordingly, thereby improving the accuracy and robustness of image preprocessing.

[0005] The first aspect of the present application provides an image correction optimization method, comprising:

[0006] determining a first distortion parameter and a first image attribute of a current character image, wherein the current character image is derived from an image acquisition device;

[0007] obtaining a historical image processing record of an image analysis device, and filtering out a secondary processing record from the historical image processing record according to the first image attribute;

[0008] generating a first correction coefficient according to the correction law of the historical correction value of automatic white balance changing with time in all secondary processing records;

[0009] determining an actual correction value of automatic white balance of the current character image, determining a predicted correction value of the current character image according to the correction law, and generating a second correction coefficient according to the deviation between the actual correction value and the predicted correction value;

[0010] The first distortion parameter is corrected according to the first correction coefficient and the second correction coefficient, and a second distortion parameter is obtained.

[0011] In some embodiments, the historical image processing records include second image attributes of historical text images, and the filtering of the secondary processing records from the historical image processing records according to the first image attributes includes:

[0012] The first image attributes are vectorized to obtain a first attribute vector, and the second image attributes are vectorized to obtain a second attribute vector;

[0013] A similarity between the first attribute vector and the second attribute vector is calculated to obtain an attribute similarity;

[0014] The historical image processing records with the attribute similarity reaching a similarity threshold are taken as the secondary processing records.

[0015] In some embodiments, the image attributes include brightness, contrast, color distribution, noise level, local spectral dispersion, and sharpening intensity, and the calculation of the similarity between the first attribute vector and the second attribute vector to obtain the attribute similarity includes:

[0016] The first attribute vector is normalized to obtain a first brightness value, a first contrast value, a first color distribution value, a first noise level value, a first local spectral dispersion value, and a first sharpening intensity value;

[0017] The second attribute vector is normalized to obtain a second brightness value, a second contrast value, a second color distribution value, a second noise level value, a second local spectral dispersion value, and a second sharpening intensity value;

[0018] Euclidean distances between the first brightness value, the first contrast value, the first color distribution value, the first noise level value, the first local spectral dispersion value, and the first sharpening intensity value, and the second brightness value, the second contrast value, the second color distribution value, the second noise level value, the second local spectral dispersion value, and the second sharpening intensity value are calculated to obtain the attribute similarity.

[0019] In some embodiments, the generation of the first correction coefficient according to a correction law of changes of historical correction values of automatic white balance over time in all the secondary processing records includes:

[0020] The historical correction values of automatic white balance in all the secondary processing records are arranged in a time stamp order to obtain a secondary correction value sequence;

[0021] The secondary correction value sequence is used to draw a correction curve of changes of automatic white balance correction values over time by statistical fitting to obtain the correction law;

[0022] Calculate the average slope of the correction curve, and take the average slope of the correction curve as a first correction coefficient.

[0023] In some embodiments, the method further comprises:

[0024] Determining a capture time point of the current character image.

[0025] Retrieving the correction law, and substituting the capture time point into the correction law to obtain a predicted correction value corresponding to the current character image.

[0026] In some embodiments, the method further comprises:

[0027] Drawing a scatter plot of the automatic white balance correction value changing over time according to the secondary correction value sequence.

[0028] Calculating the autocorrelation of the scatter plot, and determining the data characteristics according to the autocorrelation of the scatter plot.

[0029] Selecting a corresponding statistical fitting model for the automatic white balance correction value according to the data characteristics.

[0030] Solving the parameters of the statistical fitting model to obtain model parameters.

[0031] Drawing a correction curve according to the statistical fitting model and the model parameters to obtain the correction law.

[0032] In some embodiments, the method further comprises:

[0033] Weighting and calculating the first correction coefficient and the second correction coefficient to obtain a weighted value.

[0034] Adding a preset adjustment factor to the weighted value to obtain a comprehensive correction factor.

[0035] Multiplying the first distortion parameter by the comprehensive correction factor to obtain a second distortion parameter.

[0036] The second aspect of the present application provides an image correction optimization system, comprising:

[0037] A parameter attribute determination module is configured to determine a first distortion parameter and a first image attribute of a current character image, wherein the current character image is obtained from an image acquisition device.

[0038] The acquisition screening module is configured to acquire historical image processing records of the image analysis device, and screen secondary processing records from the historical image processing records according to the first image attribute;

[0039] The first coefficient generation module is configured to generate a first correction coefficient according to a correction rule of historical correction values of automatic white balance changing over time in all the secondary processing records;

[0040] The second coefficient generation module is configured to determine an actual correction value of automatic white balance of the current text image, determine a predicted correction value of the current text image according to the correction rule, and generate a second correction coefficient according to a deviation between the actual correction value and the predicted correction value.

[0041] The parameter correction module is configured to correct the first distortion parameter to obtain a second distortion parameter according to the first correction coefficient and the second correction coefficient.

[0042] The third aspect of the present application provides a computer device, comprising a memory and a processor, the memory stores a computer program, and the processor implements the steps of the above method when executing the computer program.

[0043] The fourth aspect of the present application provides a computer readable storage medium, which stores a computer program, and the computer program is executed by a processor to implement the steps of the above method.

[0044] The technical solution provided by the present application has the following advantages and effects: by comprehensively analyzing historical image processing records and real-time acquisition data, the first and second correction coefficients are dynamically generated, so that the first distortion parameter is accurately corrected. The scheme uses time series analysis and prediction of the automatic white balance correction amplitude value, effectively reflects the long-term performance degradation and immediate deviation of the target image acquisition device, makes up for the deficiency that the image analysis device cannot directly obtain the state information of the source device, greatly improves the accuracy and robustness of image preprocessing, and further improves the accuracy and stability of character recognition. The overall scheme is simple, effective and adaptable, and can effectively promote the application of image processing technology in the fields of document digitization and automatic monitoring. BRIEF DESCRIPTION OF DRAWINGS

[0045] Figure 1 is a flowchart of the image correction optimization method provided by the present application;

[0046] Figure 2 is a structural block diagram of the image correction optimization system provided by the present application;

[0047] Figure 3 is an internal structure diagram of the computer device provided by the present application. DETAILED DESCRIPTION

[0048] For the purpose of facilitating the understanding of the present application, specific embodiments of the present application will be described in more detail below with reference to the accompanying drawings.

[0049] Unless specifically stated or otherwise defined, the terms "first", "second" and the like used herein are merely used to distinguish one element from another, and do not necessarily indicate the number or order of the elements.

[0050] Unless specifically stated or otherwise defined, the term "and / or" used herein includes any and all combinations of one or more of the associated listed items.

[0051] It should be noted that "fixed to", "connected to" herein can be directly fixed or connected to an element, or indirectly fixed or connected to an element.

[0052] As shown in the embodiment, an image correction optimization method is provided, which includes the following steps S1-S5: Figure 1

[0053] Step S1, determining a first distortion parameter and a first image attribute of a current text image, wherein the current text image is from an image acquisition device.

[0054] In actual application, the target image acquisition device is responsible for collecting the original image, for example, the target image acquisition device can be a digital camera, a smartphone camera or a scanner.

[0055] Specifically, the current text image is obtained from the target image acquisition device, which refers to the image currently collected and used for text recognition processing, and the corresponding first distortion parameter refers to a set of quantitative parameters generated after preliminary evaluation of the geometric distortion existing in the text image, which is used to describe the uncertainty of the distortion mapping in the image. The set of quantitative parameters can be calculated by existing image processing algorithms. Through analysis of the current text image, the first image attribute is extracted, which is a set of quantitative characteristic parameters that can objectively reflect the image acquisition conditions and the state of the target image acquisition device, at least including: brightness, contrast, color distribution, noise level, local spectral dispersion and edge sharpness. The selected image attributes can reflect the changes of the acquisition environment and the device state to a large extent, thereby ensuring the accuracy when compared with historical data subsequently.

[0056] Step S2, obtaining a historical image processing record of an image analysis device, and screening a secondary processing record from the historical image processing record according to the first image attribute.

[0057] ​In practical applications, the image analysis device is mainly used for recognizing and processing the text in the text image. For example, the image analysis device can be a dedicated OCR server, an embedded text recognition module or a shaped intelligent view robot. The image analysis device and the image acquisition device usually exist in the form of independent modules. Since the two belong to different systems, the image analysis device cannot directly or conveniently obtain the detailed state information of the image acquisition device.

[0058] Specifically, the historical image processing records are obtained through the storage system inside the image analysis device. The historical image processing records at least include the recorded automatic white balance correction value, the time stamp, the second image attribute and other processing log information related to image correction in each image processing process. The historical image processing records provide data support for subsequent time sequence analysis to extract the performance change trend of the device.

[0059] By screening out a predetermined number of secondary processing records consistent with the current text image attribute, it can be ensured that the collection conditions of the selected historical text image and the current image are highly matched, thereby improving the accuracy of the correction coefficient.

[0060] Specifically, the secondary processing records are screened out from the historical image processing records according to the first image attribute, comprising:

[0061] The first image attribute is vectorized to obtain a first attribute vector, and the second image attribute is vectorized to obtain a second attribute vector;

[0062] The similarity between the first attribute vector and the second attribute vector is calculated to obtain an attribute similarity;

[0063] The historical image processing records with the attribute similarity reaching a similarity threshold value are taken as the secondary processing records.

[0064] In practical applications, the first brightness, the first contrast, the first color distribution, the first noise level, the first local spectral dispersion and the first sharpening intensity in the first image attribute are vectorized to obtain a first brightness vector, a first contrast vector, a first color distribution vector, a first noise level vector, a first local spectral dispersion vector and a first sharpening intensity vector. The second brightness, the second contrast, the second color distribution, the second noise level, the second local spectral dispersion and the second sharpening intensity in the second image attribute are vectorized to obtain a second brightness vector, a second contrast vector, a second color distribution vector, a second noise level vector, a second local spectral dispersion vector and a second sharpening intensity vector, so as to facilitate subsequent calculation. Through data comparison and similarity analysis, it is ensured that the selected secondary processing record is highly matched with the current text image in image acquisition conditions and device state, thereby providing a reliable data basis for subsequent time series-based trend analysis.

[0065] Specifically, the similarity between the first attribute vector and the second attribute vector is calculated to obtain an attribute similarity, including:

[0066] The first attribute vector is normalized to obtain a first brightness value, a first contrast value, a first color distribution value, a first noise level value, a first local spectral dispersion value and a first sharpening intensity value;

[0067] The second attribute vector is normalized to obtain a second brightness value, a second contrast value, a second color distribution value, a second noise level value, a second local spectral dispersion value and a second sharpening intensity value;

[0068] The Euclidean distance between the first brightness value, the first contrast value, the first color distribution value, the first noise level value, the first local spectral dispersion value and the first sharpening intensity value, and the second brightness value, the second contrast value, the second color distribution value, the second noise level value, the second local spectral dispersion value and the second sharpening intensity value is calculated to obtain an attribute similarity.

[0069] In practical applications, the first brightness vector, the first contrast vector, the first color distribution vector, the first noise level vector, the first local spectral dispersion vector and the first sharpening intensity vector are normalized to obtain a first brightness value, a first contrast value, a first color distribution value, a first noise level value, a first local spectral dispersion value and a first sharpening intensity value. The second brightness vector, the second contrast vector, the second color distribution vector, the second noise level vector, the second local spectral dispersion vector and the second sharpening intensity vector are normalized to obtain a second brightness value, a second contrast value, a second color distribution value, a second noise level value, a second local spectral dispersion value and a second sharpening intensity value. By normalizing the first attribute vector and the second attribute vector, the scale difference between the attributes is eliminated, and the calculation of the similarity is prevented from being affected by the scale difference. The attribute similarity between the first attribute vector and the second attribute vector is calculated by the Euclidean distance. The brightness value, the contrast value, the color distribution value, the noise level value, the local spectral dispersion value and the sharpening intensity value can reflect the changes of the collection environment and the device state to a large extent, thereby ensuring the accuracy of the attribute similarity between the current text image and the historical text image.

[0070] Step S3, generating a first correction coefficient according to a correction rule of the historical correction value of the automatic white balance changing over time in all secondary processing records.

[0071] In practical applications, after long-term use of the image acquisition device, the sensor and the lens assembly may have aging problems. The automatic white balance correction value, i.e., the white balance compensation amplitude, can intuitively reflect the comprehensive performance state of the sensor and the lens of the image acquisition device, including optical element aging, sensor response change, etc. When the device performance degrades, whether it is automatic compensation for color temperature or correction for image geometric distortion, a larger adjustment is needed. Therefore, the white balance compensation amplitude can be used as a comprehensive index of the "device health state" for dynamically adjusting the first distortion parameter. The first distortion parameter in the method can be calculated by using existing spatial coordinate transformation algorithms (such as perspective correction and lens distortion correction model), and the correction coefficient generated according to the historical automatic white balance correction value is only used to fine-tune the confidence interval or step size of the distortion correction model estimation. That is, instead of directly using the color correction value for geometric transformation, it is used as a reference for device performance changes to guide the geometric distortion correction algorithm to appropriately increase (or decrease) the correction strength of the potential distortion deviation during parameter optimization.

[0072] Specifically, the first correction coefficient is generated according to a correction rule of the historical correction value of the automatic white balance changing over time in all secondary processing records, including:

[0073] arranging the historical correction values of the automatic white balance in all the secondary processing records in time stamp order to obtain a secondary correction value sequence;

[0074] drawing a correction curve of the correction value of the automatic white balance changing with time by statistical fitting on the secondary correction value sequence to obtain a correction law;

[0075] calculating the average slope of the correction curve, and taking the average slope of the correction curve as a first correction coefficient.

[0076] In actual application, the correction values of the automatic white balance in all the secondary processing records are arranged in time sequence, that is, each correction value of the automatic white balance extracted from the secondary processing record is corresponded to its corresponding time stamp to obtain the secondary correction value sequence, so as to ensure the accuracy and time sequence of data sorting. Time series analysis is performed on the secondary correction value sequence, statistical fitting method is used to fit the correction value and time data, a trend curve reflecting the change of the correction value of the automatic white balance with time is drawn, and a correction curve is obtained, so that the law of automatic compensation change of the device in long-term operation is intuitively displayed. Then, the average slope of the correction curve is calculated, and the average slope represents the average change amount of the correction value of the automatic white balance per unit time, and reflects the degradation or drift trend of the device performance.

[0077] The average slope of the correction curve is taken as the first correction coefficient because the change of the device performance often leads to the change of the compensation demand of the automatic white balance, and the average slope of the correction curve quantifies the rate of change of the device performance. Taking the average slope of the correction curve as the first correction coefficient helps to dynamically compensate the influence of the degradation of the device performance when the initial distortion mapping uncertainty parameter is subsequently corrected, so as to improve the accuracy and robustness of the whole image correction process. In addition, since the image analysis device cannot directly obtain the performance change information of the image acquisition device of the processed image, the average slope of the correction curve obtained by historical data analysis can indirectly reflect the performance degradation of the image acquisition device, which provides an effective reference for subsequent dynamic correction, and further improves the stability and reliability of the image preprocessing parameter.

[0078] In step S4, the actual correction value of the automatic white balance of the current text image is determined, the predicted correction value of the current text image is determined according to the correction law, and the second correction coefficient is generated according to the deviation between the actual correction value and the predicted correction value.

[0079] Specifically, the predicted correction value of the current text image is determined according to the correction law, including:

[0080] determining the acquisition time point of the current text image;

[0081] The correction law is called, and the acquisition time point is substituted into the correction law to obtain a predicted correction value corresponding to the current character image.

[0082] In actual application, the current character image is analyzed by an image processing algorithm to determine an amplitude value of the automatic white balance correction actually applied in the current character image, i.e., an actual correction value, which can be obtained by reading parameters output by the image preprocessing module, and the acquisition time point of the image is determined by using image metadata or a system timestamp to ensure that time information required for subsequent processing is accurate.

[0083] The acquisition time point of the current character image is substituted into the correction law, i.e., a correction curve, to predict an automatic white balance correction value corresponding to the acquisition time point, i.e., a predicted correction value.

[0084] Specifically, the correction law is obtained by using statistical fitting to draw a correction curve of the automatic white balance correction value changing with time based on the secondary correction value sequence, including:

[0085] A scatter plot of the automatic white balance correction value changing with time is drawn based on the secondary correction value sequence.

[0086] Autocorrelation of the scatter plot is calculated, and data characteristics are determined based on the autocorrelation of the scatter plot.

[0087] A corresponding statistical fitting model is selected for the automatic white balance correction value based on the data characteristics.

[0088] Model parameters are obtained by solving parameters of the statistical fitting model.

[0089] A correction curve is drawn based on the statistical fitting model and the model parameters to obtain the correction law.

[0090] In actual application, a scatter plot of the automatic white balance correction value changing with time is drawn based on the secondary correction value sequence with time as the horizontal axis and the automatic white balance correction value as the vertical axis. Autocorrelation is used to measure the correlation between data at different time points. If autocorrelation coefficients present periodic changes, i.e., there are significant positive or negative values at certain fixed lags, it is determined that the data characteristics have periodicity. If autocorrelation coefficients are close to 0 at most lags, it is determined that the data characteristics have randomness. If autocorrelation coefficients slowly decay with the increase of the lag, it is determined that the data has trend. Then, a corresponding statistical fitting model is selected based on the data characteristics. For example, in the case of data having trend, a linear regression model can be selected, and the secondary correction value sequence is used to solve parameters of the linear regression model to obtain corresponding model parameters. A correction curve is drawn based on the statistical fitting model and the model parameters to obtain the correction law.

[0091] In the present application, the deviation amplitude between the actual correction value and the predicted correction value is calculated by comparing the actual correction value with the predicted correction value, the deviation amplitude reflects the difference between the actual compensation of the device and the historical trend expectation, and further reflects the possible performance abnormality or degradation of the image acquisition device at the current time. The deviation amplitude is applied as a second correction coefficient to correct the initial distortion mapping uncertainty parameter, which can further improve the accuracy and robustness of image correction on the basis of dynamically compensating the performance change of the device, and the advantage is that it can reflect the instantaneous fluctuation of the device state in real time, thereby providing more stable and reliable preprocessing parameters for the image analysis device.

[0092] Step S5, correcting the first distortion parameter according to the first correction coefficient and the second correction coefficient to obtain a second distortion parameter.

[0093] Specifically, the correcting the first distortion parameter according to the first correction coefficient and the second correction coefficient to obtain a second distortion parameter comprises:

[0094] weighting and calculating the first correction coefficient and the second correction coefficient to obtain a weighting value;

[0095] adding a preset adjustment factor to the weighting value to obtain a comprehensive correction factor;

[0096] multiplying the first distortion parameter by the comprehensive correction factor to obtain the second distortion parameter.

[0097] In actual application, the step of correcting the first distortion parameter according to the first correction coefficient and the second correction coefficient to obtain a second distortion parameter can be represented by a correction formula, and the correction formula is:

[0098]

[0099] wherein, U corrected denotes the second distortion parameter, U inital denotes the first distortion parameter, S denotes the first correction coefficient, that is, the average slope of the time-varying trend of the automatic white balance correction value, K1 denotes the adjustment weight corresponding to the first correction coefficient, A denotes the actual correction value, P denotes the predicted correction value, denotes the second correction coefficient, that is, the deviation amplitude of the actual correction value compared with the predicted correction value, K2 denotes the adjustment weight corresponding to the second correction coefficient, and 1 denotes the preset adjustment factor.

[0100] In the embodiment of the present application, by combining the first correction coefficient and the second correction coefficient to correct the first distortion parameter, the long-term degradation trend and the real-time deviation information of the image acquisition device performance can be captured at the same time, so that the final correction factor is closer to the real acquisition condition, and a more optimal correction effect is achieved.

[0101] The first correction coefficient is calculated by statistically fitting the trend of the automatic white balance correction value over time in the image processing record, and the average slope is obtained, which reflects the overall change trend of the image acquisition device performance over a long period of time; and the second correction coefficient is generated by comparing the deviation between the actual correction value of the current text image and the correction value predicted based on the historical trend, which directly reveals the immediate difference between the current image acquisition device state and the expected state.

[0102] The first correction coefficient and the second correction coefficient are closely related in the generation process: on the one hand, the first correction coefficient provides background information for long-term performance evolution, providing a stable benchmark for overall correction; on the other hand, the second correction coefficient makes up for real-time deviation, and the combination of the two can more comprehensively correct the distortion mapping uncertainty caused by device degradation or environmental changes, improve the text recognition accuracy and stability of the image analysis device, and this combination avoids the limitations of a single parameter that cannot fully reflect the device state, and realizes dynamic real-time compensation.

[0103] For example, assuming that the current text image collected by a certain image acquisition device has a first distortion parameter U corrected 0.8, and the average slope S of the trend of the automatic white balance correction value over time obtained by historical data analysis is 0.05; at the same time, the actual correction value A of the current text image is 1.2, and the predicted correction value P obtained based on the trend prediction is 1.0. Set the adjustment weight coefficients K1 and K2 to be 0.8 and 0.6 respectively. According to the formula It can be calculated that K1xS is 0.04, 0.2, 0.12, therefore, the comprehensive correction factor is 1.16, and the final first distortion parameter is 0.8 multiplied by 1.16, which is 0.928.

[0104] In this implementation process, the correction formula is called and the first distortion parameter is corrected to obtain the second distortion parameter according to the two correction coefficients, and the second distortion parameter is applied to the text recognition module of the image analysis device. As can be seen from the above example, the first correction coefficient captures the device degradation through long-term trends, and the second correction coefficient reflects the immediate deviation between the current actual and predicted values, and the combination of the two not only improves the accuracy of the correction parameter, but also enables the image analysis device to obtain more stable preprocessing parameters when facing different acquisition conditions, thereby ultimately improving the effect of text recognition.

[0105] The image correction optimization method of the present application dynamically generates first and second correction coefficients by comprehensively processing historical image processing records and real-time acquisition data, thereby accurately correcting the first distortion parameter. The scheme uses time series analysis and prediction of the automatic white balance correction amplitude value to effectively reflect the long-term performance degradation and immediate deviation of the image acquisition device, makes up for the deficiency that the image analysis device cannot directly obtain the state information of the image acquisition device, greatly improves the accuracy and robustness of image preprocessing, and further improves the accuracy and stability of character recognition. The overall scheme is simple to implement, has significant effects, and is highly adaptable, and can effectively promote the application of image processing technology in the fields of document digitization and automatic monitoring.

[0106] As shown in Figure 2 The present application embodiment also provides an image correction optimization system, which comprises:

[0107] A parameter attribute determination module 10 is configured to determine a first distortion parameter and a first image attribute of a current character image, wherein the current character image is obtained from an image acquisition device.

[0108] An acquisition and screening module 20 is configured to acquire historical image processing records of an image analysis device, and screen secondary processing records from the historical image processing records according to the first image attribute.

[0109] A first coefficient generation module 30 is configured to generate a first correction coefficient according to a correction rule of a change of a historical correction value of automatic white balance with time in all secondary processing records.

[0110] A second coefficient generation module 40 is configured to determine an actual correction value of automatic white balance of the current character image, determine a predicted correction value of the current character image according to the correction rule, and generate a second correction coefficient according to a deviation between the actual correction value and the predicted correction value.

[0111] A parameter correction module 50 is configured to correct the first distortion parameter according to the first correction coefficient and the second correction coefficient, and obtain a second distortion parameter.

[0112] The above-mentioned modules of the image correction optimization system can be realized by software, hardware, or a combination thereof. The above-mentioned modules and units can be embedded in or independent of a processor in a computer device in hardware form, or stored in a memory in a computer device in software form, so as to be called and executed by a processor to perform the operations of the above-mentioned modules.

[0113] As shown in Figure 3 The present application embodiment discloses a computer device, which comprises a memory and a processor, and the memory stores a computer program.

[0114] The computer device can be a server, and an internal structure diagram thereof can be as shown in Figure 3 The computer device includes a processor, a memory, a network interface and a database connected through a system bus. The processor of the computer device is configured to provide computing and control capabilities. The memory of the computer device includes a non-volatile storage medium and an internal memory. The non-volatile storage medium stores an operating system, a computer program and a database. The internal memory provides an environment for running the operating system and the computer program in the non-volatile storage medium. The network interface of the computer device is configured to communicate with an external terminal through a network connection. The computer program is executed by the processor to implement the image correction optimization method described in the above embodiments.

[0115] Those skilled in the art can understand that Figure 3 The structure shown in the above embodiments is only a block diagram of part of the structure related to the scheme of the present application, and does not constitute a limitation on the computer device to which the scheme of the present application is applied. Specifically, the computer device can include more or fewer components than those shown in the diagram, or combine certain components, or have a different arrangement of components.

[0116] The embodiments of the present application also disclose a computer readable storage medium storing a computer program, wherein the computer program causes a computer to execute the image correction optimization method described in the above embodiments.

[0117] Those skilled in the art can understand that all or part of the processes in the above-mentioned embodiments can be completed by a computer program instructing related hardware. The computer program can be stored in a non-volatile computer readable storage medium, and when executed, can include the processes of the above-mentioned embodiments. Any reference to a memory, storage, database or other medium used in the embodiments provided by the present application can include non-volatile and / or volatile memory. Non-volatile memory can include read-only memory (ROM), programmable ROM (PROM), electrically programmable ROM (EPROM), electrically erasable programmable ROM (EEPROM) or flash memory. Volatile memory can include random access memory (RAM) or external cache memory. As an illustration but not a limitation, RAM is available in various forms, such as static RAM (SRAM), dynamic RAM (DRAM), synchronous DRAM (SDRAM), double data rate SDRAM (DDR SDRAM), enhanced SDRAM (ESDRAM), synchronous link (Synchlink) DRAM (SLDRAM), memory bus (Rambus) direct RAM (RDRAM), direct memory bus dynamic RAM (DRDRAM) and memory bus dynamic RAM (RDRAM), etc.

[0118] The technical features of the above embodiments can be combined in any manner. For the sake of brevity, not all possible combinations of the technical features described above are described, however, it is to be understood that any combination of the technical features is within the scope of the present specification, as long as the combination is not contradictory.

Claims

1. A method of image correction optimization, characterized in that, The method comprises the following steps: determining a first distortion parameter and a first image attribute of a current text image, wherein the current text image is obtained from an image acquisition device; obtaining a historical image processing record of an image analysis device, and screening a secondary processing record from the historical image processing record according to the first image attribute; generating a first correction coefficient according to a correction law of a historical correction value of automatic white balance changing over time in all secondary processing records; determining an actual correction value of automatic white balance of the current text image, determining a predicted correction value of the current text image according to the correction law, and generating a second correction coefficient according to a deviation between the actual correction value and the predicted correction value; correcting the first distortion parameter according to the first correction coefficient and the second correction coefficient to obtain a second distortion parameter; the method of generating the first correction coefficient according to the correction law of the historical correction value of automatic white balance changing over time in all secondary processing records comprises the following steps: arranging the historical correction value of automatic white balance in all secondary processing records in a time stamp order to obtain a secondary correction value sequence; drawing a correction curve of the correction value of automatic white balance changing over time by statistical fitting on the secondary correction value sequence to obtain the correction law; calculating an average slope of the correction curve, and taking the average slope of the correction curve as the first correction coefficient; the method of correcting the first distortion parameter according to the first correction coefficient and the second correction coefficient to obtain the second distortion parameter comprises the following steps: calculating a weighted value by weighting and adding the first correction coefficient and the second correction coefficient; adding a preset adjustment factor to the weighted value to obtain a comprehensive correction factor; multiplying the first distortion parameter by the comprehensive correction factor to obtain the second distortion parameter.

2. The image correction optimization method of claim 1, wherein, the historical image processing record comprises a second image attribute of a historical text image, and the method of screening the secondary processing record from the historical image processing record according to the first image attribute comprises the following steps: vectorizing the first image attribute to obtain a first attribute vector, and vectorizing the second image attribute to obtain a second attribute vector; calculating a similarity between the first attribute vector and the second attribute vector to obtain an attribute similarity; taking the historical image processing record with the attribute similarity reaching a similarity threshold as the secondary processing record.

3. The image correction optimization method of claim 2, wherein, The image attribute comprises brightness, contrast, color distribution, noise level, local spectral dispersion degree and sharpening intensity, and the method of calculating the similarity between the first attribute vector and the second attribute vector to obtain the attribute similarity comprises the following steps: normalizing the first attribute vector to obtain a first brightness value, a first contrast value, a first color distribution value, a first noise level value, a first local spectral dispersion degree value and a first sharpening intensity value; normalizing the second attribute vector to obtain a second brightness value, a second contrast value, a second color distribution value, a second noise level value, a second local spectral dispersion degree value and a second sharpening intensity value; calculating a Euclidean distance between the first luminance value, the first contrast value, the first color distribution value, the first noise level value, the first local spectral dispersion value and the first sharpening intensity value, and the second luminance value, the second contrast value, the second color distribution value, the second noise level value, the second local spectral dispersion value and the second sharpening intensity value, to obtain an attribute similarity.

4. The image correction optimization method of claim 1, wherein, The method further comprises: determining a collection time point of the current character image; substituting the collection time point into the correction law to obtain a predicted correction value corresponding to the current character image.

5. The image correction optimization method of claim 1, wherein, The method further comprises: drawing a scatter plot of the automatic white balance correction value changing over time according to the secondary correction value sequence; calculating an autocorrelation of the scatter plot, and determining a data characteristic according to the autocorrelation of the scatter plot; selecting a corresponding statistical fitting model for the automatic white balance correction value according to the data characteristic; solving parameters of the statistical fitting model to obtain model parameters; drawing a correction curve according to the statistical fitting model and the model parameters to obtain the correction law.

6. An image correction optimization system characterized by, The method further comprises: a parameter attribute determination module configured to determine a first distortion parameter and a first image attribute of a current character image, wherein the current character image is obtained from an image acquisition device; an acquisition and screening module configured to acquire a historical image processing record of an image analysis device, and screen a secondary processing record from the historical image processing record according to the first image attribute; a first coefficient generation module configured to generate a first correction coefficient according to a correction law of a historical correction value of automatic white balance changing over time in all secondary processing records; a second coefficient generation module configured to determine an actual correction value of automatic white balance of the current character image, determine a predicted correction value of the current character image according to the correction law, and generate a second correction coefficient according to a deviation between the actual correction value and the predicted correction value; a parameter correction module configured to correct the first distortion parameter according to the first correction coefficient and the second correction coefficient to obtain a second distortion parameter. The first coefficient generation module is specifically configured to arrange the historical correction value of automatic white balance in all secondary processing records in a time stamp order to obtain a secondary correction value sequence. The method further comprises: drawing a correction curve of the automatic white balance correction value changing over time according to the secondary correction value sequence to obtain the correction law; calculating an average slope of the correction curve, and taking the average slope of the correction curve as the first correction coefficient; The parameter correction module is specifically configured to weight and calculate the first correction coefficient and the second correction coefficient to obtain a weighted value. adding a preset adjustment factor to the weighted value to obtain a comprehensive correction factor; 7. A computer device comprising a memory and a processor, the memory storing a computer program, characterized in that, multiplying the first distortion parameter by the comprehensive correction factor to obtain the second distortion parameter.

8. A computer readable storage medium having stored thereon a computer program, characterized in that The computer program is executed by the processor to implement the steps of the method of any one of claims 1-5. The computer program is executed by the processor to implement the steps of the method of any one of claims 1-5.

Images