CA3140466A1 - Image quality determination method, apparatus, and system - Google Patents
Image quality determination method, apparatus, and system Download PDFInfo
- Publication number
- CA3140466A1 CA3140466A1 CA3140466A CA3140466A CA3140466A1 CA 3140466 A1 CA3140466 A1 CA 3140466A1 CA 3140466 A CA3140466 A CA 3140466A CA 3140466 A CA3140466 A CA 3140466A CA 3140466 A1 CA3140466 A1 CA 3140466A1
- Authority
- CA
- Canada
- Prior art keywords
- indicator
- image
- image quality
- edge pixels
- determining
- 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
- 238000000034 method Methods 0.000 title claims abstract description 45
- 238000004364 calculation method Methods 0.000 claims abstract description 33
- 238000010801 machine learning Methods 0.000 claims description 22
- 230000035945 sensitivity Effects 0.000 claims description 8
- 230000009466 transformation Effects 0.000 claims description 7
- 238000012549 training Methods 0.000 claims description 6
- 238000006243 chemical reaction Methods 0.000 claims description 3
- 230000006872 improvement Effects 0.000 claims description 3
- 238000004891 communication Methods 0.000 description 6
- 238000012545 processing Methods 0.000 description 6
- 238000010586 diagram Methods 0.000 description 3
- 238000001514 detection method Methods 0.000 description 2
- 230000000694 effects Effects 0.000 description 2
- 238000007726 management method Methods 0.000 description 2
- 238000007781 pre-processing Methods 0.000 description 2
- 230000008569 process Effects 0.000 description 2
- 238000013139 quantization Methods 0.000 description 2
- 241000699670 Mus sp. Species 0.000 description 1
- 230000009286 beneficial effect Effects 0.000 description 1
- 238000013500 data storage Methods 0.000 description 1
- 238000011161 development Methods 0.000 description 1
- 238000005516 engineering process Methods 0.000 description 1
- 230000006870 function Effects 0.000 description 1
- 238000003702 image correction Methods 0.000 description 1
- 230000003993 interaction Effects 0.000 description 1
- 230000003287 optical effect Effects 0.000 description 1
- 230000000750 progressive effect Effects 0.000 description 1
- 230000003068 static effect Effects 0.000 description 1
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T7/00—Image analysis
- G06T7/0002—Inspection of images, e.g. flaw detection
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/24—Classification techniques
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N20/00—Machine learning
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V10/00—Arrangements for image or video recognition or understanding
- G06V10/20—Image preprocessing
- G06V10/22—Image preprocessing by selection of a specific region containing or referencing a pattern; Locating or processing of specific regions to guide the detection or recognition
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T2207/00—Indexing scheme for image analysis or image enhancement
- G06T2207/30—Subject of image; Context of image processing
- G06T2207/30168—Image quality inspection
Landscapes
- Engineering & Computer Science (AREA)
- Theoretical Computer Science (AREA)
- General Physics & Mathematics (AREA)
- Physics & Mathematics (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Data Mining & Analysis (AREA)
- Software Systems (AREA)
- General Engineering & Computer Science (AREA)
- Evolutionary Computation (AREA)
- Artificial Intelligence (AREA)
- Multimedia (AREA)
- Bioinformatics & Cheminformatics (AREA)
- Bioinformatics & Computational Biology (AREA)
- Evolutionary Biology (AREA)
- Medical Informatics (AREA)
- Life Sciences & Earth Sciences (AREA)
- Computing Systems (AREA)
- Mathematical Physics (AREA)
- Quality & Reliability (AREA)
- Image Analysis (AREA)
Abstract
The present invention discloses to an image quality determination method, apparatus, and system. The method comprises: converting a received license document image into a grayscale image; determining edge pixels of the grayscale image, respectively judging whether all edge pixels with Roof structure in the edge pixels meet blur condition, calculating number of edge pixels with Roof structure that meet the blur condition, and a ratio of this number of edge pixels to number of all edge pixels with Roof structure is an indicator of image blurriness; calculating an indicator of image texture noise according to pre-set first calculation rule; calculating an indicator of image contrast according to pre-set second calculation rule;
determining image quality by using the indicator of blurriness, the indicator of texture noise and the indicator of contrast. The present application can accurately determine image quality of license document comparing with prior art.
determining image quality by using the indicator of blurriness, the indicator of texture noise and the indicator of contrast. The present application can accurately determine image quality of license document comparing with prior art.
Description
IMAGE QUALITY DETERMINATION METHOD, APPARATUS, AND SYSTEM
Field [0001] The present disclosure relates to image processing field, particularly to an image quality determination method, apparatus, and system.
Background
Field [0001] The present disclosure relates to image processing field, particularly to an image quality determination method, apparatus, and system.
Background
[0002] In financial business, such as when a company applies for a loan form a financial institution, the company business license needs to be provided to enter the system for subsequent risk management procedures. Using automatic machinery identification instead of manually entry can greatly reduce entry costs and improve entry efficiency.
[0003] After the existing system performs image correction, red badge removing and other pre-processing, then recognizing the image content, the pre-processing can improve the image quality which the original image is slanted and with red badge, thereby effectively improving the recognition accuracy. However, image quality suffers from the influence of miscellaneous factors, blur caused by out of focus, paper texture, excessive light and other factors will significantly reduce the image quality and make the image content difficult to accurately recognize.
[0004] In order to reduce the recognition errors caused by poor image quality , before using the recognition system to recognize the image content, firstly analyzing of the merits of image quality, do not perform processing on lower quality image, prompting the user to re-upload, therefore, the accuracy of image quality determination is very important, and the methods for determining image quality in the prior art are generalized image quality determination, although the image quality of the license document can also be determined, the image quality of the license document cannot be accurately determined.
Date recue / Date received 202 1-1 1-25 Invention Content
Date recue / Date received 202 1-1 1-25 Invention Content
[0005] The present application provides an image quality determination method, apparatus and system which can accurately determine the image quality of license document.
[0006] The present application provides the following solutions:
[0007] The first aspect provides an image quality determination method, the method comprises:
[0008] Converting a received license document image into a grayscale image;
[0009] Determining edge pixels of the grayscale image, respectively judging whether all edge pixels with Roof structure in the edge pixels meet blur condition, calculating number of edge pixels with Roof structure, wherein the edge pixels meet the blur condition, and a ratio of the number of this edge pixels to the number of all edge pixels with Roof structure is an indicator of image blurriness;
[0010] Calculating an indicator of image texture noise according to pre-set first calculation rule;
[0011] Calculating an indicator of image contrast according to pre-set second calculation rule;
[0012] Determining image quality by using the indicator of blurriness, the indicator of texture noise and the indicator of contrast.
[0013] Furthermore, wherein the determination of edge pixels of the grayscale image comprises:
[0014] Performing a third-order wavelet transform on the grayscale image, extracting low frequency components obtained by wavelet transform of each order;
Date recue / Date received 202 1-1 1-25
Date recue / Date received 202 1-1 1-25
[0015] Determining edge pixels according to pre-set rule by using three low frequency components.
[0016] Furthermore, wherein calculating an indicator of image texture noise according to pre-set first calculation rule comprises:
[0017] When performing the third-order wavelet transformation on the grayscale image, extracting diagonal direction component H Hi, horizontal direction component HLi and vertical direction component LHi obtained by performing first-order wavelet transformation on the grayscale image;
[0018] Calculating the texture noise indicator 8,, by using following formula:
max[median(IHH11),medianaHL11),median(ILH11)]
[ 0019] 8 =
0.6745 [0020] Furthermore, wherein calculating an indicator of image contrast according to pre-set second calculation rule comprises:
[0021] After dividing the grayscale image into k1 x k2 blocks, respectively calculating standard deviation of each block by using the following formula:
E7_21(ii-02 [0022] a =
n2 [0023] Wherein, a is intensity standard deviation of all pixels in block, n is number of pixels in block, refers to intensity of each pixel in block, I refers to average intensity of all pixels in block;
Date recue / Date received 202 1-1 1-25 [0024] Comparing the standard deviation of each block with a pre-set sensitivity threshold, the blocks with less standard deviation than the pre-set sensitivity threshold are filtered to delete;
[0025] Calculating contrast indicator RME for the remaining blocks by using following formular:
11+12: +inl [0026] RME =
k k1k2 i=1 1 loglIti-1-11+12+
[0027] Wherein, RME refers to roots' improvement, k1 and k2 respectively refers to number of blocks in each row and each column; i and j respectively refers to X-axis and Y-axis of block in grayscale image;
refers to pixel intensity at midpoint of block; n refers to number of pixels in each block, and /2 in refers to intensity of each pixel in block.
[0028] Furthermore, determining image quality by using the indicator of blurriness, the indicator of texture noise and the indicator of contrast comprises:
[0029] Judging whether the indicator of blurriness is greater than a first threshold, whether the indicator of texture noise is greater than a second threshold, and whether the indicator of contrast is less than the third threshold, if yes, determining the image as low quality, if not, then determining the image as high quality.
[0030] Determining image quality by using the indicator of blurriness, the indicator of texture noise and the indicator of contrast comprises:
[0031] Inputting the indicator of blurriness, the indicator of texture noise, and the indicator of the contrast as variables into a pre-trained first machine learning model, wherein a training set in this pre-trained first machine learning model is a collection of image sample data marked with image quality;
Date recue / Date received 202 1-1 1-25 [0032] Determining the image quality by the pre-trained first machine learning model.
[0033] Furthermore, determining image quality by using the indicator of blurriness, the indicator of texture noise and the indicator of contrast comprises:
[0034] Inputting the indicator of blurriness, the indicator of texture noise, and the indicator of the contrast as variables into a pre-trained second machine learning model, wherein a training set in this pre-trained second machine learning model is a collection of image sample data marked with OCR accurate rate;
[0035] Calculating the estimated value of the OCR accurate rate of the image according to the pre-trained second machine learning model, if the estimated value of OCR accurate rate is greater than a pre-set fourth threshold , determining the image as high quality, otherwise, determining the image as low quality.
[0036] Furthermore, the method also comprises:
[0037] Transmitting images to OCR recognition system for recognizing, wherein the images meet quality requirements.
[0038] The second aspect of the present application provides an image quality determination apparatus, comprising:
[0039] An image conversion unit configured to convert the received license document image into a grayscale image;
[0040] A first calculating unit configured to determine edge pixels of the grayscale image, respectively judging whether all edge pixels with Roof structure in the edge pixels meet blur condition, calculating edge pixel numbers with Roof structure, wherein the edge pixel meets the blur condition, and a ratio of this edge Date recue / Date received 202 1-1 1-25 pixel numbers to all edge pixels numbers with Roof structure is an indicator of image blurriness;
[0041] A second calculation unit configured to calculate an indicator of image texture noise according to pre-set first calculation rule;
[0042] A third calculation unit configured to an indicator of image contrast according to pre-set second calculation rule;
[0043] A determining unit configured determining image quality by using the indicator of blurriness, the indicator of texture noise and the indicator of contrast.
[0044] The third aspect of the present application provides a computer system, comprising:
[0045] One or plural processors; and [0046] A memory associated with one or plural processors, the memory is configured to store program commands, if the program commands are executed by one or plural processors, executing any above-mentioned method.
[0047] According to the specific implementation provided in this application, this application discloses the following technical effects: the ratio of the number of edge pixels of the conditional Roof structure to the number of edge pixels of all the Roof structure is taken as the indicator of blurriness, because most of the content of the license document images are text content, and the strength of both sides of the text edge is the same, most of the edge pixels of the words are Roof structure, and only considering whether the edge pixels of the Roof structure meet the blur condition, which accurately judge the blurriness of the words font; calculating the indicator of the texture noise according to the pre-set first calculation rule; calculating the indicator of contrast according to the pre-set second calculation rule;
and the image quality can be Date recue / Date received 202 1-1 1-25 determined more accurately by using the indicator of blurriness, the indicator of texture noise, and the indicator of contrast to determine the image quality.
Drawing Description [0048] In order to describe the technical solutions clearer in the implementations of the present application or the prior art, the following are drawings that need to be used are briefly introduced.
Obviously, the drawings in the following description are only some implementations of the application, for those of ordinary skill in the art , without creative work, they can also obtain other drawings based on these drawings.
[0049] Figure 1 is a process diagram of an image quality determination method in implementation 1 of the present application;
[0050] Figure 2 is a structural diagram of an image quality determination apparatus in implementation 2 of the present application;
[0051] Figure 3 is a structural diagram of computer system in implementation 3 of the present application.
Specific implementation methods [0052] The following will describe the technical solutions of the implementations in the present application with accompanying drawings, obviously the described implementations are only a part of the implementations in the present application. Based on the implementations in the present application, all other implementations obtained by those of ordinary skilled in the art will fall in the protection scope of the present application.
Date recue / Date received 202 1-1 1-25 [0053] As the above-mentioned in the technical background, in order to reduce the recognition errors caused by poor image quality , before using OCR to recognize the image content, firstly analyzing of the merits of image quality, do not perform processing on lower quality image, prompting the user to re-upload, therefore, the accuracy of image quality determination is very important, and the methods for determining image quality in the prior art are generalized image quality determination, although the image quality of the license document can also be determined, it is not targeted and the image quality of the license document cannot be accurately determined.
[0054] The present application provides an image quality determination method with calculating the ratio of the number of edge pixels of the conditional Roof structure to the number of edge pixels of all the Roof structure, the ratio is taken as the indicator of blurriness, because most of the content of the license document images are text content, and the strength of both sides of the text edge is the same, most of the edge pixels of the words are Roof structure, and only considering whether the edge pixels of the Roof structure meet the blur condition, which accurately judge the blurriness of the words font;
calculating the indicator of the texture noise according to the pre-set first calculation rule; calculating the indicator of contrast according to the pre-set second calculation rule; and the image quality can be determined more accurately by using the indicator of blurriness, the indicator of texture noise, and the indicator of contrast to determine the image quality.
[0055] Implementation 1 [0056] The present application provides an image quality determination method, and the method is applied to an image quality determination apparatus as an example, the apparatus can be configured in any computer device, so that the computer device can execute image quality determination method.
[0057] As shown in Figure 1, the above-mentioned method comprises:
Date recue / Date received 202 1-1 1-25 [0058] S11, converting a received license document image into a grayscale image;
[0059] S12, determining edge pixels of the grayscale image, respectively judging whether all edge pixels with Roof structure in the edge pixels meet blur condition, calculating number of edge pixels with Roof structure, wherein the edge pixels meet the blur condition, and a ratio of the number of this edge pixels to the number of all edge pixels with Roof structure is an indicator of image blurriness;
[0060]
There are four types of edge pixels: Dirac structure, Astep structure, Gstep structure and Roof structure, due to most of the image content of the license document are text content, and the strength of both sides of the text edge are the same, so the edge of the text pixels are mostly with Roof structure, only considering whether the edge pixels of the Roof structure meet the blur condition, which can be more accurately determining the font blurriness, so calculating a ratio of the number of edge pixels with the Roof structure that meets the blurriness condition to the number of all edge pixels with Roof structure, this ratio is taken to make image blun-iness quantization, as an indicator of the image blurriness, wherein the determination of whether the edge pixels with Roof structure meet the blur condition is based on whether the strongness of the edge pixels in the first-order low-frequency component is less than the pre-set threshold which can determine whether it is blurry.
[0061] S13, calculating an indicator of image texture noise according to pre-set first calculation rule;
[0062] S14, calculating an indicator of image contrast according to pre-set second calculation rule;
[0063] Because image quality suffers from the influence of miscellaneous factors, blur caused by out of focus, paper texture, excessive light and other factors will significantly reduce the image quality, therefore, when determining image quality, comprehensively considering is required to respectively calculate the indicator of texture noise and contrast.
Date recue / Date received 202 1-1 1-25 [0064] S15, determining image quality by using the indicator of blurriness, the indicator of texture noise and the indicator of contrast.
[0065] The determination of the edge pixels of the grayscale image comprises:
[0066] Performing a third-order wavelet transform on the grayscale image, extracting low frequency components obtained by wavelet transform of each order;
[0067] Determining edge pixels according to pre-set rule by using three low frequency components.
[0068] Determining edge pixels of a grayscale image, it is necessary to perform a third-order wavelet transform on the grayscale image, each wavelet transformation will obtain one low-frequency component, performing third-order wavelet transform will obtain three low-frequency components, determining the edge pixels by using the three low-frequency components according to pre-set rule.
[0069] Calculating an indicator of image texture noise according to pre-set first calculation rule comprises:
[0070] When performing the third-order wavelet transformation on the grayscale image, extracting diagonal direction component H Hi, horizontal direction component HLi and vertical direction component LHi obtained by performing first-order wavelet transformation on the grayscale image;
[0071] Calculating the texture noise indicator 6,, by using following formula:
0072] = max[median(IH Hil),median(IHL11),median(ILHi 0]
[ 8 0.6745 [0073] Due to factors such as paper folding, moisture or camera wave pattern, a series of texture noises Date recue / Date received 202 1-1 1-25 will be generated in the image. These texture noises often block text content, directly affect the detection of text content by the detection model, and seriously affect OCR accuracy rate, because the texture noise of the image of the license document is directional , the texture noise on the diagonal, the horizontal and the vertical directions of the image need to be considered comprehensively , taking in the strongest three texture noise on the three direction as the image texture noise, it can be more accurately determining the image quality, so taking the maximum value of the median absolute value among the diagonal direction component HHi , horizontal direction component HLi and vertical direction component LHi when calculation.
[0074] Calculating an indicator of image contrast according to pre-set second calculation rule comprises:
[0075] After dividing the grayscale image into k1 x k2 blocks, respectively calculating standard deviation of each block by using the following formula:
n2 E =
[0076] a = _ n2 [0077] Wherein, a is intensity standard deviation of all pixels in block, n is number of pixels in block, refers to intensity of each pixel in block, I refers to average intensity of all pixels in block;
[0078] Comparing the standard deviation of each block with a pre-set sensitivity threshold, the blocks with less standard deviation than the pre-set sensitivity threshold are filtered to delete;
[0079] Calculating contrast indicator RME for the remaining blocks by using following formular:
11+12: +inl [0080] RME =
k kik2 i=1 loglIti+11+12+ 11 Date recue / Date received 202 1-1 1-25 [0081] Wherein, RME(Root Mean Enhance) refers to roots' improvement, k1 and k2 respectively refers to number of blocks in each row and each column; i and j respectively refers to X-axis and Y-axis of block in grayscale image; I refers to pixel intensity at midpoint of block; n refers to number of pixels in each block, and 12 In refers to intensity of each pixel in block.
[0082] In a low-contrast document image, the text is generally lighter, and the distinction between the paper background is low, and it can generally be detected correctly, but it will affect the accuracy of recognition. For the image of license documents, the part without text content often have no big difference, and the contrast of this part does not need to be considered, so after dividing the grayscale image into k1 x k2 blocks , calculating the standard deviation of each block, comparing the standard deviation of each block with a pre-set sensitivity threshold, the blocks with less standard deviation than the pre-set sensitivity threshold are filtered to delete, regardless of the contrast of this part which can filter out the pure paper background part and focus on calculating the contrast between the text and the background to more accurately determine the image quality.
[0083] Determining image quality by using the indicator of blurriness, the indicator of texture noise and the indicator of contrast comprises:
[0084] Judging whether the indicator of blurriness is greater than a first threshold, whether the indicator of texture noise is greater than a second threshold, and whether the indicator of contrast is less than the third threshold, if yes, determining the image as low quality, if not, then determining the image as high quality.
[0085] When determining the image quality, respectively determining the indicator of blurriness, the indicator of texture noise and the indicator of contrast, and the indicator of the first threshold value, the second threshold value and the third threshold value, which are suitable for large amount of data and no marked image source.
Date recue / Date received 202 1-1 1-25 [0086] Determining image quality by using the indicator of blurriness, the indicator of texture noise and the indicator of contrast comprises:
[0087] Inputting the indicator of blurriness, the indicator of texture noise, and the indicator of the contrast as variables into a pre-trained first machine learning model, wherein a training set in this pre-trained first machine learning model is a collection of image sample data marked with image quality;
[0088] Determining the image quality by the pre-trained first machine learning model.
[0089] When image quality is judged by threshold, only a single indicator is considered each time, which will cause inaccuracy of the determination of image quality, for example, an image has a certain level of texture noise , blurriness and low contrast at the same time, but which have not reached the threshold, the image will be determined to be high quality , but not easy to recognize, for OCR recognition system, the image is low quality, therefore, at the beginning of the development of the OCR recognition system, by marking the image as 'high quality' and low quality' manually to transform the problem to the machine learning's binary classification task which can quickly and accurately determine the image quality.
[0090] Determining image quality by using the indicator of blurriness, the indicator of texture noise and the indicator of contrast comprises:
[0091] Inputting the indicator of blurriness, the indicator of texture noise, and the indicator of the contrast as variables into a pre-trained second machine learning model, wherein a training set in this pre-trained second machine learning model is a collection of image sample data marked with OCR accurate rate;
[0092] Calculating the estimated value of the OCR accurate rate of the image according to the pre-trained second machine learning model, if the estimated value of OCR accurate rate is greater than a pre-set fourth Date recue / Date received 202 1-1 1-25 threshold , determining the image as high quality, otherwise, determining the image as low quality.
[0093] When manually marking image as 'high quality' and low quality', there will be problem of subjective judgments by observers, when a poor-quality image appears after a poor-quality image, usually will be marked as 'high quality', in addition, some low-quality features for people may not affect the recognition effect of the OCR recognition system, such as slanted angle, old papers and so on. Therefore, the OCR accuracy rate can be used as a mark to convert the problem into machine learning's OCR accuracy rate prediction, the OCR recognition system recognizes the image, the image has been marked with the correct text content, and calculating the accurate rate of OCR by comparing the recognized text content and marked text content, regarding the OCR accurate rate as a tag image of the sample, using the image sample data set that already marked with OCR accurate rate to train the second machine learning model, using the trained second machine learning model to calculate the OCR accurate rate of the image, the higher OCR accurate rate, the higher image quality.
[0094] The method also comprises:
[0095] Transmitting images to OCR recognition system for recognizing, wherein the images meet quality requirements.
[0096] Implementation 2 [0097] Corresponding to the above-mentioned method, the present application provides an image quality determination apparatus, as shown in Figure 2, comprising:
[0098] An image conversion unit 21 configured to convert the received license document image into a grayscale image;
Date recue / Date received 202 1-1 1-25 [0099]
A first calculating unit 22 configured to determine edge pixels of the grayscale image, respectively judging whether all edge pixels with Roof structure in the edge pixels meet blur condition, calculating edge pixel numbers with Roof structure, wherein the edge pixel meets the blur condition, and a ratio of this edge pixel numbers to all edge pixels numbers with Roof structure is an indicator of image blurriness;
[0100]
There are four types of edge pixels: Dirac structure, Astep structure, Gstep structure and Roof structure, due to most of the image content of the license document are text content, and the strength of both sides of the text edge are the same, so the edge of the text pixels are mostly with Roof structure, only considering whether the edge pixels of the Roof structure meet the blur condition, which can be more accurately determining the font blurriness, so calculating a ratio of the number of edge pixels with the Roof structure that meets the blurriness condition to the number of all edge pixels with Roof structure, this ratio is taken to make image blurriness quantization, as the indicator of image blurriness.
[0101] A second calculation unit 23 configured to calculate an indicator of image texture noise according to pre-set first calculation rule;
[0102] A third calculation unit 24 configured to an indicator of image contrast according to pre-set second calculation rule;
[0103] Because image quality suffers from the influence of miscellaneous factors, blur caused by out of focus, paper texture, excessive light and other factors will significantly reduce the image quality, therefore, when determining image quality, comprehensively considering is required to respectively calculate the indicator of texture noise and contrast.
[0104] A determining unit 25 configured determining image quality by using the indicator of blurriness, the indicator of texture noise and the indicator of contrast.
Date recue / Date received 202 1-1 1-25 [0105] The implementation of the present application provides an image quality determining apparatus, which belongs to the same application concept with the image quality method provided by the implementation of the present application, the image quality determination method provided in the implementation of this application can be executed, which will have correspondingly functional modules and beneficial effects of this image quality determination method. For the technical details that are not described in this implementation can refer to the image quality determination method provided in the implementation of this application, which will not be repeated here.
[0106] Implementation 3 [0107] Corresponding to the above-mentioned method and apparatus, the implementation 3 of the present application provides a computer system, comprising:
[0108] One or plural processors; and [0109] A memory associated with one or plural processors, the memory is configured to store program commands, if the program commands are executed by one or plural processors, executing the method step in implementation 1, such as following steps:
[0110] Converting a received license document image into a grayscale image;
[0111] Determining edge pixels of the grayscale image, respectively judging whether all edge pixels with Roof structure in the edge pixels meet blur condition, calculating number of edge pixels with Roof structure, wherein the edge pixels meet the blur condition, and a ratio of the number of this edge pixels to the number of all edge pixels with Roof structure is an indicator of image blurriness;
Date recue / Date received 202 1-1 1-25 [0112] Calculating an indicator of image texture noise according to pre-set first calculation rule;
[0113] Calculating an indicator of image contrast according to pre-set second calculation rule;
[0114] Determining image quality by using the indicator of blurriness, the indicator of texture noise and the indicator of contrast.
[0115] Wherein, Figure 3 exemplarily shows the architecture of the computer system , which can specifically include a processor 1510, video display adapter 1511, disk driver 1512, input/output interface 1513, network interface 1514, and memory 1520. The above-mentioned processor 1510, video display adapter 1511, disk driver 1512, input/output interface 1513, network interface 1514 and memory 1520 can be connected through a communication bus 1530.
[0116] Wherein, the processor 1510 can be achieved by using a general CPU
(Central Processing Unit), Microprocessor, Application Specific Integrated Circuit (ASIC) , or one or more integrated circuits, which are used to execute some relative program to achieve the technical solutions provided in this application.
[0117]
The memory 1520 can adopt ROM ( Read Only Memory), RAM (Random Access Memory), static storage devices and dynamic storage devices to achieve. The memory 1520 can store operate system 1521 used to control the running of the computer system 1500, used to control the low-level operation of the computer system 1500's Basic Input Output System (BIOS) 1522.
In addition, storing a web browser 1523, data storage management 1524, and icon font processing system 1525 and so on. The above-mentioned icon font processing system 1525 can be the specific application that implements the above-mentioned steps. To sum up, when achieving the technical solutions provided by this application through software or firmware, related program codes are stored in the memory 1520 and executed by a processor 1510.
Date recue / Date received 202 1-1 1-25 [0118] Input / output interface 1513 is used for connecting input / output modules to achieve the information input and output. Input / output module can be configured in the device as a component ( not shown in the figure), or it can be connected to the device to provide corresponding functions. Wherein, Input devices can include keyboards, mice, touch screens, microphones, various sensors, etc., and output devices can include monitors, speakers, vibrators, lights and so on.
[0119] The network interface 1514 is used to connect a communication module (not shown in the figure) to achieve the communication interaction between this device and other devices. Wherein, the communication module can achieve communication through wired means (such as USB , network cable, etc. ) , or through wireless methods ( such as mobile network, WIFI, Bluetooth, etc.) to achieve communication.
[0120] The bus 1530 includes a path and transmits information among various components of the device (such as the processor 1510 , the video display adapter 1511, the disk driver 1512, the input/output interface 1513, the network interface 1514, and the memory 1520).
[0121] In addition, the computer system 1500 also can obtain information with specific receiving conditions from the virtual resource object's receiving condition information database 1541 for condition judgement and son on.
[0122] It should be noted that although the above device only shows the processor 1510, the video display adapter 1511, the disk driver 1512, input/output interface 1513, network interface 1514, memory 1520, bus 1530, etc., but in the process of the specific implementation, the device may also include other essential components for normal operation. In addition, those skilled in the art can understand that the above apparatus can comprise only the essential components of the present application to achieve the implementation, but there is no need to contain all the components as shown in figure.
Date recue / Date received 202 1-1 1-25 [O123] Known from the description of the above implementations that those skilled in the art can clearly understand that the application can be achieved with the help of software and essential general hardware platform. Based on this understanding, the essence of the technical solution of this application, or in other words, the part that contributes to the existing technology can be implemented in the form of a software product, the computer software product can be stored in storage media, such as ROM/RAM, magnetic disks, optical disks, etc., including several commands to make a computer device (can be a personal computer, a cloud server, or a network device, etc.) to execute the methods described in each implementation or some of the implementations of the present application.
[0124] The various implementations in this description are described in a progressive manner, the same and similar parts among the various implementations can be referred to each other separately, and each implementation focuses on the differences compared with the other implementations. Especially for the concern of the system or the system implementations, since it is basically similar to the implementation method, the description is relatively simple. For related details, please refer to the implementation method. The system and system implementations described in the above are only illustrative, and the units described by separate parts may or may not be physically separate, and the parts displayed as units may or may not be physical units, which means, it can be in one place, or it may be distributed to plural network units. Some or all the modules are selected according to actual needs to achieve the implementation's solution purpose. The ordinary skill in the art can understand and implement without creative work.
[0125] The image quality determination method, apparatus, and system provided by this application are described in detail in the above. Specific examples are used to illustrate the principle and implementation of this application. The description of the above examples is only for helping to understand the methods and core ideas of this application; at the same time, for those of ordinary skill in the art, according to the ideas of this application, there will be changes in the specific implementations and the scopes. In summary, the content of this description should not be taken to be the restrictions of the present application.
max[median(IHH11),medianaHL11),median(ILH11)]
[ 0019] 8 =
0.6745 [0020] Furthermore, wherein calculating an indicator of image contrast according to pre-set second calculation rule comprises:
[0021] After dividing the grayscale image into k1 x k2 blocks, respectively calculating standard deviation of each block by using the following formula:
E7_21(ii-02 [0022] a =
n2 [0023] Wherein, a is intensity standard deviation of all pixels in block, n is number of pixels in block, refers to intensity of each pixel in block, I refers to average intensity of all pixels in block;
Date recue / Date received 202 1-1 1-25 [0024] Comparing the standard deviation of each block with a pre-set sensitivity threshold, the blocks with less standard deviation than the pre-set sensitivity threshold are filtered to delete;
[0025] Calculating contrast indicator RME for the remaining blocks by using following formular:
11+12: +inl [0026] RME =
k k1k2 i=1 1 loglIti-1-11+12+
[0027] Wherein, RME refers to roots' improvement, k1 and k2 respectively refers to number of blocks in each row and each column; i and j respectively refers to X-axis and Y-axis of block in grayscale image;
refers to pixel intensity at midpoint of block; n refers to number of pixels in each block, and /2 in refers to intensity of each pixel in block.
[0028] Furthermore, determining image quality by using the indicator of blurriness, the indicator of texture noise and the indicator of contrast comprises:
[0029] Judging whether the indicator of blurriness is greater than a first threshold, whether the indicator of texture noise is greater than a second threshold, and whether the indicator of contrast is less than the third threshold, if yes, determining the image as low quality, if not, then determining the image as high quality.
[0030] Determining image quality by using the indicator of blurriness, the indicator of texture noise and the indicator of contrast comprises:
[0031] Inputting the indicator of blurriness, the indicator of texture noise, and the indicator of the contrast as variables into a pre-trained first machine learning model, wherein a training set in this pre-trained first machine learning model is a collection of image sample data marked with image quality;
Date recue / Date received 202 1-1 1-25 [0032] Determining the image quality by the pre-trained first machine learning model.
[0033] Furthermore, determining image quality by using the indicator of blurriness, the indicator of texture noise and the indicator of contrast comprises:
[0034] Inputting the indicator of blurriness, the indicator of texture noise, and the indicator of the contrast as variables into a pre-trained second machine learning model, wherein a training set in this pre-trained second machine learning model is a collection of image sample data marked with OCR accurate rate;
[0035] Calculating the estimated value of the OCR accurate rate of the image according to the pre-trained second machine learning model, if the estimated value of OCR accurate rate is greater than a pre-set fourth threshold , determining the image as high quality, otherwise, determining the image as low quality.
[0036] Furthermore, the method also comprises:
[0037] Transmitting images to OCR recognition system for recognizing, wherein the images meet quality requirements.
[0038] The second aspect of the present application provides an image quality determination apparatus, comprising:
[0039] An image conversion unit configured to convert the received license document image into a grayscale image;
[0040] A first calculating unit configured to determine edge pixels of the grayscale image, respectively judging whether all edge pixels with Roof structure in the edge pixels meet blur condition, calculating edge pixel numbers with Roof structure, wherein the edge pixel meets the blur condition, and a ratio of this edge Date recue / Date received 202 1-1 1-25 pixel numbers to all edge pixels numbers with Roof structure is an indicator of image blurriness;
[0041] A second calculation unit configured to calculate an indicator of image texture noise according to pre-set first calculation rule;
[0042] A third calculation unit configured to an indicator of image contrast according to pre-set second calculation rule;
[0043] A determining unit configured determining image quality by using the indicator of blurriness, the indicator of texture noise and the indicator of contrast.
[0044] The third aspect of the present application provides a computer system, comprising:
[0045] One or plural processors; and [0046] A memory associated with one or plural processors, the memory is configured to store program commands, if the program commands are executed by one or plural processors, executing any above-mentioned method.
[0047] According to the specific implementation provided in this application, this application discloses the following technical effects: the ratio of the number of edge pixels of the conditional Roof structure to the number of edge pixels of all the Roof structure is taken as the indicator of blurriness, because most of the content of the license document images are text content, and the strength of both sides of the text edge is the same, most of the edge pixels of the words are Roof structure, and only considering whether the edge pixels of the Roof structure meet the blur condition, which accurately judge the blurriness of the words font; calculating the indicator of the texture noise according to the pre-set first calculation rule; calculating the indicator of contrast according to the pre-set second calculation rule;
and the image quality can be Date recue / Date received 202 1-1 1-25 determined more accurately by using the indicator of blurriness, the indicator of texture noise, and the indicator of contrast to determine the image quality.
Drawing Description [0048] In order to describe the technical solutions clearer in the implementations of the present application or the prior art, the following are drawings that need to be used are briefly introduced.
Obviously, the drawings in the following description are only some implementations of the application, for those of ordinary skill in the art , without creative work, they can also obtain other drawings based on these drawings.
[0049] Figure 1 is a process diagram of an image quality determination method in implementation 1 of the present application;
[0050] Figure 2 is a structural diagram of an image quality determination apparatus in implementation 2 of the present application;
[0051] Figure 3 is a structural diagram of computer system in implementation 3 of the present application.
Specific implementation methods [0052] The following will describe the technical solutions of the implementations in the present application with accompanying drawings, obviously the described implementations are only a part of the implementations in the present application. Based on the implementations in the present application, all other implementations obtained by those of ordinary skilled in the art will fall in the protection scope of the present application.
Date recue / Date received 202 1-1 1-25 [0053] As the above-mentioned in the technical background, in order to reduce the recognition errors caused by poor image quality , before using OCR to recognize the image content, firstly analyzing of the merits of image quality, do not perform processing on lower quality image, prompting the user to re-upload, therefore, the accuracy of image quality determination is very important, and the methods for determining image quality in the prior art are generalized image quality determination, although the image quality of the license document can also be determined, it is not targeted and the image quality of the license document cannot be accurately determined.
[0054] The present application provides an image quality determination method with calculating the ratio of the number of edge pixels of the conditional Roof structure to the number of edge pixels of all the Roof structure, the ratio is taken as the indicator of blurriness, because most of the content of the license document images are text content, and the strength of both sides of the text edge is the same, most of the edge pixels of the words are Roof structure, and only considering whether the edge pixels of the Roof structure meet the blur condition, which accurately judge the blurriness of the words font;
calculating the indicator of the texture noise according to the pre-set first calculation rule; calculating the indicator of contrast according to the pre-set second calculation rule; and the image quality can be determined more accurately by using the indicator of blurriness, the indicator of texture noise, and the indicator of contrast to determine the image quality.
[0055] Implementation 1 [0056] The present application provides an image quality determination method, and the method is applied to an image quality determination apparatus as an example, the apparatus can be configured in any computer device, so that the computer device can execute image quality determination method.
[0057] As shown in Figure 1, the above-mentioned method comprises:
Date recue / Date received 202 1-1 1-25 [0058] S11, converting a received license document image into a grayscale image;
[0059] S12, determining edge pixels of the grayscale image, respectively judging whether all edge pixels with Roof structure in the edge pixels meet blur condition, calculating number of edge pixels with Roof structure, wherein the edge pixels meet the blur condition, and a ratio of the number of this edge pixels to the number of all edge pixels with Roof structure is an indicator of image blurriness;
[0060]
There are four types of edge pixels: Dirac structure, Astep structure, Gstep structure and Roof structure, due to most of the image content of the license document are text content, and the strength of both sides of the text edge are the same, so the edge of the text pixels are mostly with Roof structure, only considering whether the edge pixels of the Roof structure meet the blur condition, which can be more accurately determining the font blurriness, so calculating a ratio of the number of edge pixels with the Roof structure that meets the blurriness condition to the number of all edge pixels with Roof structure, this ratio is taken to make image blun-iness quantization, as an indicator of the image blurriness, wherein the determination of whether the edge pixels with Roof structure meet the blur condition is based on whether the strongness of the edge pixels in the first-order low-frequency component is less than the pre-set threshold which can determine whether it is blurry.
[0061] S13, calculating an indicator of image texture noise according to pre-set first calculation rule;
[0062] S14, calculating an indicator of image contrast according to pre-set second calculation rule;
[0063] Because image quality suffers from the influence of miscellaneous factors, blur caused by out of focus, paper texture, excessive light and other factors will significantly reduce the image quality, therefore, when determining image quality, comprehensively considering is required to respectively calculate the indicator of texture noise and contrast.
Date recue / Date received 202 1-1 1-25 [0064] S15, determining image quality by using the indicator of blurriness, the indicator of texture noise and the indicator of contrast.
[0065] The determination of the edge pixels of the grayscale image comprises:
[0066] Performing a third-order wavelet transform on the grayscale image, extracting low frequency components obtained by wavelet transform of each order;
[0067] Determining edge pixels according to pre-set rule by using three low frequency components.
[0068] Determining edge pixels of a grayscale image, it is necessary to perform a third-order wavelet transform on the grayscale image, each wavelet transformation will obtain one low-frequency component, performing third-order wavelet transform will obtain three low-frequency components, determining the edge pixels by using the three low-frequency components according to pre-set rule.
[0069] Calculating an indicator of image texture noise according to pre-set first calculation rule comprises:
[0070] When performing the third-order wavelet transformation on the grayscale image, extracting diagonal direction component H Hi, horizontal direction component HLi and vertical direction component LHi obtained by performing first-order wavelet transformation on the grayscale image;
[0071] Calculating the texture noise indicator 6,, by using following formula:
0072] = max[median(IH Hil),median(IHL11),median(ILHi 0]
[ 8 0.6745 [0073] Due to factors such as paper folding, moisture or camera wave pattern, a series of texture noises Date recue / Date received 202 1-1 1-25 will be generated in the image. These texture noises often block text content, directly affect the detection of text content by the detection model, and seriously affect OCR accuracy rate, because the texture noise of the image of the license document is directional , the texture noise on the diagonal, the horizontal and the vertical directions of the image need to be considered comprehensively , taking in the strongest three texture noise on the three direction as the image texture noise, it can be more accurately determining the image quality, so taking the maximum value of the median absolute value among the diagonal direction component HHi , horizontal direction component HLi and vertical direction component LHi when calculation.
[0074] Calculating an indicator of image contrast according to pre-set second calculation rule comprises:
[0075] After dividing the grayscale image into k1 x k2 blocks, respectively calculating standard deviation of each block by using the following formula:
n2 E =
[0076] a = _ n2 [0077] Wherein, a is intensity standard deviation of all pixels in block, n is number of pixels in block, refers to intensity of each pixel in block, I refers to average intensity of all pixels in block;
[0078] Comparing the standard deviation of each block with a pre-set sensitivity threshold, the blocks with less standard deviation than the pre-set sensitivity threshold are filtered to delete;
[0079] Calculating contrast indicator RME for the remaining blocks by using following formular:
11+12: +inl [0080] RME =
k kik2 i=1 loglIti+11+12+ 11 Date recue / Date received 202 1-1 1-25 [0081] Wherein, RME(Root Mean Enhance) refers to roots' improvement, k1 and k2 respectively refers to number of blocks in each row and each column; i and j respectively refers to X-axis and Y-axis of block in grayscale image; I refers to pixel intensity at midpoint of block; n refers to number of pixels in each block, and 12 In refers to intensity of each pixel in block.
[0082] In a low-contrast document image, the text is generally lighter, and the distinction between the paper background is low, and it can generally be detected correctly, but it will affect the accuracy of recognition. For the image of license documents, the part without text content often have no big difference, and the contrast of this part does not need to be considered, so after dividing the grayscale image into k1 x k2 blocks , calculating the standard deviation of each block, comparing the standard deviation of each block with a pre-set sensitivity threshold, the blocks with less standard deviation than the pre-set sensitivity threshold are filtered to delete, regardless of the contrast of this part which can filter out the pure paper background part and focus on calculating the contrast between the text and the background to more accurately determine the image quality.
[0083] Determining image quality by using the indicator of blurriness, the indicator of texture noise and the indicator of contrast comprises:
[0084] Judging whether the indicator of blurriness is greater than a first threshold, whether the indicator of texture noise is greater than a second threshold, and whether the indicator of contrast is less than the third threshold, if yes, determining the image as low quality, if not, then determining the image as high quality.
[0085] When determining the image quality, respectively determining the indicator of blurriness, the indicator of texture noise and the indicator of contrast, and the indicator of the first threshold value, the second threshold value and the third threshold value, which are suitable for large amount of data and no marked image source.
Date recue / Date received 202 1-1 1-25 [0086] Determining image quality by using the indicator of blurriness, the indicator of texture noise and the indicator of contrast comprises:
[0087] Inputting the indicator of blurriness, the indicator of texture noise, and the indicator of the contrast as variables into a pre-trained first machine learning model, wherein a training set in this pre-trained first machine learning model is a collection of image sample data marked with image quality;
[0088] Determining the image quality by the pre-trained first machine learning model.
[0089] When image quality is judged by threshold, only a single indicator is considered each time, which will cause inaccuracy of the determination of image quality, for example, an image has a certain level of texture noise , blurriness and low contrast at the same time, but which have not reached the threshold, the image will be determined to be high quality , but not easy to recognize, for OCR recognition system, the image is low quality, therefore, at the beginning of the development of the OCR recognition system, by marking the image as 'high quality' and low quality' manually to transform the problem to the machine learning's binary classification task which can quickly and accurately determine the image quality.
[0090] Determining image quality by using the indicator of blurriness, the indicator of texture noise and the indicator of contrast comprises:
[0091] Inputting the indicator of blurriness, the indicator of texture noise, and the indicator of the contrast as variables into a pre-trained second machine learning model, wherein a training set in this pre-trained second machine learning model is a collection of image sample data marked with OCR accurate rate;
[0092] Calculating the estimated value of the OCR accurate rate of the image according to the pre-trained second machine learning model, if the estimated value of OCR accurate rate is greater than a pre-set fourth Date recue / Date received 202 1-1 1-25 threshold , determining the image as high quality, otherwise, determining the image as low quality.
[0093] When manually marking image as 'high quality' and low quality', there will be problem of subjective judgments by observers, when a poor-quality image appears after a poor-quality image, usually will be marked as 'high quality', in addition, some low-quality features for people may not affect the recognition effect of the OCR recognition system, such as slanted angle, old papers and so on. Therefore, the OCR accuracy rate can be used as a mark to convert the problem into machine learning's OCR accuracy rate prediction, the OCR recognition system recognizes the image, the image has been marked with the correct text content, and calculating the accurate rate of OCR by comparing the recognized text content and marked text content, regarding the OCR accurate rate as a tag image of the sample, using the image sample data set that already marked with OCR accurate rate to train the second machine learning model, using the trained second machine learning model to calculate the OCR accurate rate of the image, the higher OCR accurate rate, the higher image quality.
[0094] The method also comprises:
[0095] Transmitting images to OCR recognition system for recognizing, wherein the images meet quality requirements.
[0096] Implementation 2 [0097] Corresponding to the above-mentioned method, the present application provides an image quality determination apparatus, as shown in Figure 2, comprising:
[0098] An image conversion unit 21 configured to convert the received license document image into a grayscale image;
Date recue / Date received 202 1-1 1-25 [0099]
A first calculating unit 22 configured to determine edge pixels of the grayscale image, respectively judging whether all edge pixels with Roof structure in the edge pixels meet blur condition, calculating edge pixel numbers with Roof structure, wherein the edge pixel meets the blur condition, and a ratio of this edge pixel numbers to all edge pixels numbers with Roof structure is an indicator of image blurriness;
[0100]
There are four types of edge pixels: Dirac structure, Astep structure, Gstep structure and Roof structure, due to most of the image content of the license document are text content, and the strength of both sides of the text edge are the same, so the edge of the text pixels are mostly with Roof structure, only considering whether the edge pixels of the Roof structure meet the blur condition, which can be more accurately determining the font blurriness, so calculating a ratio of the number of edge pixels with the Roof structure that meets the blurriness condition to the number of all edge pixels with Roof structure, this ratio is taken to make image blurriness quantization, as the indicator of image blurriness.
[0101] A second calculation unit 23 configured to calculate an indicator of image texture noise according to pre-set first calculation rule;
[0102] A third calculation unit 24 configured to an indicator of image contrast according to pre-set second calculation rule;
[0103] Because image quality suffers from the influence of miscellaneous factors, blur caused by out of focus, paper texture, excessive light and other factors will significantly reduce the image quality, therefore, when determining image quality, comprehensively considering is required to respectively calculate the indicator of texture noise and contrast.
[0104] A determining unit 25 configured determining image quality by using the indicator of blurriness, the indicator of texture noise and the indicator of contrast.
Date recue / Date received 202 1-1 1-25 [0105] The implementation of the present application provides an image quality determining apparatus, which belongs to the same application concept with the image quality method provided by the implementation of the present application, the image quality determination method provided in the implementation of this application can be executed, which will have correspondingly functional modules and beneficial effects of this image quality determination method. For the technical details that are not described in this implementation can refer to the image quality determination method provided in the implementation of this application, which will not be repeated here.
[0106] Implementation 3 [0107] Corresponding to the above-mentioned method and apparatus, the implementation 3 of the present application provides a computer system, comprising:
[0108] One or plural processors; and [0109] A memory associated with one or plural processors, the memory is configured to store program commands, if the program commands are executed by one or plural processors, executing the method step in implementation 1, such as following steps:
[0110] Converting a received license document image into a grayscale image;
[0111] Determining edge pixels of the grayscale image, respectively judging whether all edge pixels with Roof structure in the edge pixels meet blur condition, calculating number of edge pixels with Roof structure, wherein the edge pixels meet the blur condition, and a ratio of the number of this edge pixels to the number of all edge pixels with Roof structure is an indicator of image blurriness;
Date recue / Date received 202 1-1 1-25 [0112] Calculating an indicator of image texture noise according to pre-set first calculation rule;
[0113] Calculating an indicator of image contrast according to pre-set second calculation rule;
[0114] Determining image quality by using the indicator of blurriness, the indicator of texture noise and the indicator of contrast.
[0115] Wherein, Figure 3 exemplarily shows the architecture of the computer system , which can specifically include a processor 1510, video display adapter 1511, disk driver 1512, input/output interface 1513, network interface 1514, and memory 1520. The above-mentioned processor 1510, video display adapter 1511, disk driver 1512, input/output interface 1513, network interface 1514 and memory 1520 can be connected through a communication bus 1530.
[0116] Wherein, the processor 1510 can be achieved by using a general CPU
(Central Processing Unit), Microprocessor, Application Specific Integrated Circuit (ASIC) , or one or more integrated circuits, which are used to execute some relative program to achieve the technical solutions provided in this application.
[0117]
The memory 1520 can adopt ROM ( Read Only Memory), RAM (Random Access Memory), static storage devices and dynamic storage devices to achieve. The memory 1520 can store operate system 1521 used to control the running of the computer system 1500, used to control the low-level operation of the computer system 1500's Basic Input Output System (BIOS) 1522.
In addition, storing a web browser 1523, data storage management 1524, and icon font processing system 1525 and so on. The above-mentioned icon font processing system 1525 can be the specific application that implements the above-mentioned steps. To sum up, when achieving the technical solutions provided by this application through software or firmware, related program codes are stored in the memory 1520 and executed by a processor 1510.
Date recue / Date received 202 1-1 1-25 [0118] Input / output interface 1513 is used for connecting input / output modules to achieve the information input and output. Input / output module can be configured in the device as a component ( not shown in the figure), or it can be connected to the device to provide corresponding functions. Wherein, Input devices can include keyboards, mice, touch screens, microphones, various sensors, etc., and output devices can include monitors, speakers, vibrators, lights and so on.
[0119] The network interface 1514 is used to connect a communication module (not shown in the figure) to achieve the communication interaction between this device and other devices. Wherein, the communication module can achieve communication through wired means (such as USB , network cable, etc. ) , or through wireless methods ( such as mobile network, WIFI, Bluetooth, etc.) to achieve communication.
[0120] The bus 1530 includes a path and transmits information among various components of the device (such as the processor 1510 , the video display adapter 1511, the disk driver 1512, the input/output interface 1513, the network interface 1514, and the memory 1520).
[0121] In addition, the computer system 1500 also can obtain information with specific receiving conditions from the virtual resource object's receiving condition information database 1541 for condition judgement and son on.
[0122] It should be noted that although the above device only shows the processor 1510, the video display adapter 1511, the disk driver 1512, input/output interface 1513, network interface 1514, memory 1520, bus 1530, etc., but in the process of the specific implementation, the device may also include other essential components for normal operation. In addition, those skilled in the art can understand that the above apparatus can comprise only the essential components of the present application to achieve the implementation, but there is no need to contain all the components as shown in figure.
Date recue / Date received 202 1-1 1-25 [O123] Known from the description of the above implementations that those skilled in the art can clearly understand that the application can be achieved with the help of software and essential general hardware platform. Based on this understanding, the essence of the technical solution of this application, or in other words, the part that contributes to the existing technology can be implemented in the form of a software product, the computer software product can be stored in storage media, such as ROM/RAM, magnetic disks, optical disks, etc., including several commands to make a computer device (can be a personal computer, a cloud server, or a network device, etc.) to execute the methods described in each implementation or some of the implementations of the present application.
[0124] The various implementations in this description are described in a progressive manner, the same and similar parts among the various implementations can be referred to each other separately, and each implementation focuses on the differences compared with the other implementations. Especially for the concern of the system or the system implementations, since it is basically similar to the implementation method, the description is relatively simple. For related details, please refer to the implementation method. The system and system implementations described in the above are only illustrative, and the units described by separate parts may or may not be physically separate, and the parts displayed as units may or may not be physical units, which means, it can be in one place, or it may be distributed to plural network units. Some or all the modules are selected according to actual needs to achieve the implementation's solution purpose. The ordinary skill in the art can understand and implement without creative work.
[0125] The image quality determination method, apparatus, and system provided by this application are described in detail in the above. Specific examples are used to illustrate the principle and implementation of this application. The description of the above examples is only for helping to understand the methods and core ideas of this application; at the same time, for those of ordinary skill in the art, according to the ideas of this application, there will be changes in the specific implementations and the scopes. In summary, the content of this description should not be taken to be the restrictions of the present application.
19 Date recue / Date received 202 1-1 1-25
Claims (10)
1. An image quality determination method comprises:
converting a received license document image into a grayscale image;
determining edge pixels of the grayscale image, respectively judging whether all edge pixels with Roof structure in the edge pixels meet blur condition, calculating number of edge pixels with Roof structure, wherein the edge pixels meet the blur condition, and a ratio of the number of this edge pixels to the number of all edge pixels with Roof structure is an indicator of image blurriness;
calculating an indicator of image texture noise according to pre-set first calculation rule;
calculating an indicator of image contrast according to pre-set second calculation rule; and determining image quality by using the indicator of blurriness, the indicator of texture noise and the indicator of contrast.
converting a received license document image into a grayscale image;
determining edge pixels of the grayscale image, respectively judging whether all edge pixels with Roof structure in the edge pixels meet blur condition, calculating number of edge pixels with Roof structure, wherein the edge pixels meet the blur condition, and a ratio of the number of this edge pixels to the number of all edge pixels with Roof structure is an indicator of image blurriness;
calculating an indicator of image texture noise according to pre-set first calculation rule;
calculating an indicator of image contrast according to pre-set second calculation rule; and determining image quality by using the indicator of blurriness, the indicator of texture noise and the indicator of contrast.
2. The image quality determination method according to claim 1, wherein the determination of edge pixels of the grayscale image comprises:
performing a third-order wavelet transform on the grayscale image, extracting low frequency components obtained by wavelet transform of each order; and determining edge pixels according to pre-set rule by using three low frequency components.
performing a third-order wavelet transform on the grayscale image, extracting low frequency components obtained by wavelet transform of each order; and determining edge pixels according to pre-set rule by using three low frequency components.
3. The image quality determination method according to claim 2, wherein calculating an indicator of image texture noise according to pre-set first calculation rule comprises:
Date recue / Date received 2021-11-25 when performing the third-order wavelet transformation on the grayscale image, extracting diagonal direction component HH1, horizontal direction component HL1 and vertical direction component LH1 obtained by performing first-order wavelet transformation on the grayscale image;
calculating the texture noise indicator ô,. by using following formula:
max[median(lHH11),median(lHL11),median(ILH101 = _____________________________________________________________ 0.6745
Date recue / Date received 2021-11-25 when performing the third-order wavelet transformation on the grayscale image, extracting diagonal direction component HH1, horizontal direction component HL1 and vertical direction component LH1 obtained by performing first-order wavelet transformation on the grayscale image;
calculating the texture noise indicator ô,. by using following formula:
max[median(lHH11),median(lHL11),median(ILH101 = _____________________________________________________________ 0.6745
4. The image quality determination method according to claim 1, wherein calculating an indicator of image contrast according to pre-set second calculation rule comprises:
after dividing the grayscale image into k1 x k2 blocks, respectively calculating standard deviation of each block by using the following formula:
ni 21 T)2 = _____________________________________________ n2 wherein, a is intensity standard deviation of all pixels in block, n is number of pixels in block, refers to intensity of each pixel in block, I refers to average intensity of all pixels in block;
comparing the standard deviation of each block with a pre-set sensitivity threshold, the blocks with less standard deviation than the pre-set sensitivity threshold are filtered to delete;
calculating contrast indicator R1VIE for the remaining blocks by using following fonnular:
RME=
kik2 i=1 L'i=1 log bij + /2 +...
wherein, RME refers to roots' improvement, k1 and k2 respectively refers to number of blocks Date recue / Date received 2021-11-25 in each row and each column; i and j respectively refers to X-axis and Y-axis of block in grayscale image; Iii refers to pixel intensity at midpoint of block; n refers to number of pixels in each block, and 11, 4, refers to intensity of each pixel in block.
after dividing the grayscale image into k1 x k2 blocks, respectively calculating standard deviation of each block by using the following formula:
ni 21 T)2 = _____________________________________________ n2 wherein, a is intensity standard deviation of all pixels in block, n is number of pixels in block, refers to intensity of each pixel in block, I refers to average intensity of all pixels in block;
comparing the standard deviation of each block with a pre-set sensitivity threshold, the blocks with less standard deviation than the pre-set sensitivity threshold are filtered to delete;
calculating contrast indicator R1VIE for the remaining blocks by using following fonnular:
RME=
kik2 i=1 L'i=1 log bij + /2 +...
wherein, RME refers to roots' improvement, k1 and k2 respectively refers to number of blocks Date recue / Date received 2021-11-25 in each row and each column; i and j respectively refers to X-axis and Y-axis of block in grayscale image; Iii refers to pixel intensity at midpoint of block; n refers to number of pixels in each block, and 11, 4, refers to intensity of each pixel in block.
5. Any image quality determination method according to claim 1 to 4, determining image quality by using the indicator of blurriness, the indicator of texture noise and the indicator of contrast comprises:
judging whether the indicator of blurriness is greater than a first threshold, whether the indicator of texture noise is greater than a second threshold, and whether the indicator of contrast is less than the third threshold, if yes, determining the image as low quality, if not, then determining the image as high quality.
judging whether the indicator of blurriness is greater than a first threshold, whether the indicator of texture noise is greater than a second threshold, and whether the indicator of contrast is less than the third threshold, if yes, determining the image as low quality, if not, then determining the image as high quality.
6. Any image quality determination method according to claim 1 to 4, determining image quality by using the indicator of blurriness, the indicator of texture noise and the indicator of contrast comprises:
inputting the indicator of blurriness, the indicator of texture noise, and the indicator of the contrast as variables into a pre-trained first machine learning model, wherein a training set in this pre-trained first machine learning model is a collection of image sample data marked with image quality; and obtaining the result of determining the image quality by the pre-trained first machine learning model.
inputting the indicator of blurriness, the indicator of texture noise, and the indicator of the contrast as variables into a pre-trained first machine learning model, wherein a training set in this pre-trained first machine learning model is a collection of image sample data marked with image quality; and obtaining the result of determining the image quality by the pre-trained first machine learning model.
7. Any image quality determination method according to claim 1 to 4, determining image quality by using the indicator of blurriness, the indicator of texture noise and the indicator of contrast comprises:
inputting the indicator of blurriness, the indicator of texture noise, and the indicator of the contrast as variables into a pre-trained second machine learning model, wherein a training set in this pre-Date recue / Date received 2021-11-25 trained second machine learning model is a collection of image sample data marked with OCR
accurate rate; and calculating the estimated value of the OCR accurate rate of the image according to the pre-trained second machine learning model, if the estimated value of OCR accurate rate is greater than a pre-set fourth threshold , determining the image as high quality, otherwise, determining the image as low quality.
inputting the indicator of blurriness, the indicator of texture noise, and the indicator of the contrast as variables into a pre-trained second machine learning model, wherein a training set in this pre-Date recue / Date received 2021-11-25 trained second machine learning model is a collection of image sample data marked with OCR
accurate rate; and calculating the estimated value of the OCR accurate rate of the image according to the pre-trained second machine learning model, if the estimated value of OCR accurate rate is greater than a pre-set fourth threshold , determining the image as high quality, otherwise, determining the image as low quality.
8. The image quality determination method according to claim 1, comprising:
transmitting images to OCR recognition system for recognizing, wherein the images meet quality requirements.
transmitting images to OCR recognition system for recognizing, wherein the images meet quality requirements.
9. An image quality determination apparatus comprises:
an image conversion unit configured to convert the received license document image into a gray scale image;
a first calculating unit configured to determine edge pixels of the grayscale image, respectively judging whether all edge pixels with Roof structure in the edge pixels meet blur condition, calculating edge pixel numbers with Roof structure, wherein the edge pixel meets the blur condition, and a ratio of this edge pixel numbers to all edge pixels numbers with Roof structure is an indicator of image blurriness;
a second calculation unit configured to calculate an indicator of image texture noise according to pre-set first calculation rule;
Date recue / Date received 2021-11-25 a third calculation unit configured to an indicator of image contrast according to pre-set second calculation rule; and A determining unit configured determining image quality by using the indicator of blurriness, the indicator of texture noise and the indicator of contrast.
an image conversion unit configured to convert the received license document image into a gray scale image;
a first calculating unit configured to determine edge pixels of the grayscale image, respectively judging whether all edge pixels with Roof structure in the edge pixels meet blur condition, calculating edge pixel numbers with Roof structure, wherein the edge pixel meets the blur condition, and a ratio of this edge pixel numbers to all edge pixels numbers with Roof structure is an indicator of image blurriness;
a second calculation unit configured to calculate an indicator of image texture noise according to pre-set first calculation rule;
Date recue / Date received 2021-11-25 a third calculation unit configured to an indicator of image contrast according to pre-set second calculation rule; and A determining unit configured determining image quality by using the indicator of blurriness, the indicator of texture noise and the indicator of contrast.
10. A computer system comprises:
one or plural processors; and a memory associated with one or plural processors, the memory is configured to store program commands, if the program commands are executed by one or plural processors, executing any method in claim 1 to 8.
Date recue / Date received 2021-11-25
one or plural processors; and a memory associated with one or plural processors, the memory is configured to store program commands, if the program commands are executed by one or plural processors, executing any method in claim 1 to 8.
Date recue / Date received 2021-11-25
Applications Claiming Priority (2)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202011337307.2 | 2020-11-25 | ||
| CN202011337307.2A CN112419275B (en) | 2020-11-25 | 2020-11-25 | Image quality determination method, device and system |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CA3140466A1 true CA3140466A1 (en) | 2022-05-25 |
Family
ID=74842077
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CA3140466A Pending CA3140466A1 (en) | 2020-11-25 | 2021-11-25 | Image quality determination method, apparatus, and system |
Country Status (2)
| Country | Link |
|---|---|
| CN (1) | CN112419275B (en) |
| CA (1) | CA3140466A1 (en) |
Families Citing this family (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN115870815B (en) * | 2022-12-28 | 2023-06-16 | 中山超精科技有限公司 | Lens polishing detection method and device |
| CN117676038B (en) * | 2024-01-30 | 2024-04-05 | 北京点聚信息技术有限公司 | Electronic license data secure sharing method and system |
Family Cites Families (6)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN106856002B (en) * | 2016-11-22 | 2020-06-26 | 上海大学 | Unmanned aerial vehicle shooting image quality evaluation method |
| CN106960436A (en) * | 2017-03-20 | 2017-07-18 | 哈尔滨工业大学 | A kind of remote sensing image process performance appraisal procedure |
| CN107784651B (en) * | 2017-10-31 | 2021-03-02 | 中国矿业大学 | Fuzzy image quality evaluation method based on fuzzy detection weighting |
| JP7185477B2 (en) * | 2018-10-12 | 2022-12-07 | 株式会社Pfu | Image processing device, control method and control program |
| CN110232680B (en) * | 2019-05-30 | 2021-04-27 | 广智微芯(扬州)有限公司 | Image ambiguity evaluation method and device |
| CN110443783B (en) * | 2019-07-08 | 2021-10-15 | 新华三信息安全技术有限公司 | Image quality evaluation method and device |
-
2020
- 2020-11-25 CN CN202011337307.2A patent/CN112419275B/en active Active
-
2021
- 2021-11-25 CA CA3140466A patent/CA3140466A1/en active Pending
Also Published As
| Publication number | Publication date |
|---|---|
| CN112419275B (en) | 2023-03-10 |
| CN112419275A (en) | 2021-02-26 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| US20210117705A1 (en) | Traffic image recognition method and apparatus, and computer device and medium | |
| US8867828B2 (en) | Text region detection system and method | |
| CA3140466A1 (en) | Image quality determination method, apparatus, and system | |
| US12118714B2 (en) | Method of detecting and classifying defects and electronic device using the same | |
| US9679218B2 (en) | Method and apparatus for image matching | |
| CN114387591A (en) | License plate recognition method, system, equipment and storage medium | |
| CN112784835B (en) | Method and device for identifying authenticity of circular seal, electronic equipment and storage medium | |
| CN112085721A (en) | Damage assessment method, device and equipment for flooded vehicle based on artificial intelligence and storage medium | |
| CN109145723A (en) | A kind of seal recognition methods, system, terminal installation and storage medium | |
| CN116311214A (en) | License plate recognition method and device | |
| US20100111439A1 (en) | Noise Reduction For Digital Images | |
| CN106663212B (en) | Character recognition device, character recognition method, and computer-readable storage medium | |
| CN111199240A (en) | Training method of bank card identification model, and bank card identification method and device | |
| CN112101139B (en) | Human shape detection method, device, equipment and storage medium | |
| CN112749735A (en) | Converter tapping steel flow identification method, system, medium and terminal based on deep learning | |
| CN116665146A (en) | Vehicle thatch cover detection method and device, electronic equipment and storage medium | |
| CN111369531A (en) | Image definition grading method, equipment and storage device | |
| CN116935368A (en) | Deep learning model training method, text line detection method, device and equipment | |
| CN114612919B (en) | Bill information processing system, method and device | |
| CN111178340B (en) | Image recognition method and training method of image recognition model | |
| TWI775038B (en) | Method and device for recognizing character and storage medium | |
| CN114821596A (en) | Text recognition method and device, electronic equipment and medium | |
| CN111797922B (en) | Text image classification method and device | |
| CN110166767B (en) | Image quality adjusting method and device combining vehicle machine idle resources | |
| CN110263782B (en) | Card detection method and device based on deep learning |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| EEER | Examination request |
Effective date: 20220916 |
|
| EEER | Examination request |
Effective date: 20220916 |
|
| EEER | Examination request |
Effective date: 20220916 |
|
| EEER | Examination request |
Effective date: 20220916 |
|
| EEER | Examination request |
Effective date: 20220916 |
|
| EEER | Examination request |
Effective date: 20220916 |
|
| EEER | Examination request |
Effective date: 20220916 |
|
| EEER | Examination request |
Effective date: 20220916 |
|
| EEER | Examination request |
Effective date: 20220916 |