CN114863492A - Method and device for repairing low-quality fingerprint image - Google Patents
Method and device for repairing low-quality fingerprint image Download PDFInfo
- Publication number
- CN114863492A CN114863492A CN202210785657.8A CN202210785657A CN114863492A CN 114863492 A CN114863492 A CN 114863492A CN 202210785657 A CN202210785657 A CN 202210785657A CN 114863492 A CN114863492 A CN 114863492A
- Authority
- CN
- China
- Prior art keywords
- image
- point
- continuity
- line
- filtering
- 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.)
- Granted
Links
- 238000000034 method Methods 0.000 title claims abstract description 70
- 238000001914 filtration Methods 0.000 claims abstract description 71
- 238000012545 processing Methods 0.000 claims abstract description 56
- 238000004364 calculation method Methods 0.000 claims abstract description 22
- 230000008439 repair process Effects 0.000 claims abstract description 19
- 230000002093 peripheral effect Effects 0.000 claims abstract description 5
- 230000008569 process Effects 0.000 claims description 32
- 230000002146 bilateral effect Effects 0.000 claims description 24
- 230000008859 change Effects 0.000 claims description 21
- 238000000605 extraction Methods 0.000 claims description 8
- 238000005070 sampling Methods 0.000 claims description 8
- 206010040925 Skin striae Diseases 0.000 claims description 7
- 230000002708 enhancing effect Effects 0.000 claims description 5
- 238000011156 evaluation Methods 0.000 claims description 5
- 230000011218 segmentation Effects 0.000 claims description 5
- 239000000126 substance Substances 0.000 claims description 3
- 230000009286 beneficial effect Effects 0.000 abstract description 4
- 238000012937 correction Methods 0.000 description 18
- 238000010586 diagram Methods 0.000 description 11
- 230000000694 effects Effects 0.000 description 9
- 230000001965 increasing effect Effects 0.000 description 8
- 238000007429 general method Methods 0.000 description 7
- 238000013519 translation Methods 0.000 description 6
- 238000009792 diffusion process Methods 0.000 description 3
- 238000013459 approach Methods 0.000 description 2
- 230000007547 defect Effects 0.000 description 2
- 238000009499 grossing Methods 0.000 description 2
- 210000001624 hip Anatomy 0.000 description 2
- 239000011159 matrix material Substances 0.000 description 2
- 238000003672 processing method Methods 0.000 description 2
- 230000007704 transition Effects 0.000 description 2
- 230000001154 acute effect Effects 0.000 description 1
- 238000004458 analytical method Methods 0.000 description 1
- 239000003086 colorant Substances 0.000 description 1
- 238000001514 detection method Methods 0.000 description 1
- 208000037265 diseases, disorders, signs and symptoms Diseases 0.000 description 1
- 239000006185 dispersion Substances 0.000 description 1
- 238000005516 engineering process Methods 0.000 description 1
- 238000012986 modification Methods 0.000 description 1
- 230000004048 modification Effects 0.000 description 1
- 238000000926 separation method Methods 0.000 description 1
Images
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V40/00—Recognition of biometric, human-related or animal-related patterns in image or video data
- G06V40/10—Human or animal bodies, e.g. vehicle occupants or pedestrians; Body parts, e.g. hands
- G06V40/12—Fingerprints or palmprints
- G06V40/1347—Preprocessing; Feature extraction
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T3/00—Geometric image transformation in the plane of the image
- G06T3/40—Scaling the whole image or part thereof
- G06T3/4007—Interpolation-based scaling, e.g. bilinear interpolation
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T5/00—Image enhancement or restoration
- G06T5/20—Image enhancement or restoration by the use of local operators
-
- G06T5/70—
-
- G06T5/73—
-
- G06T5/77—
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T2207/00—Indexing scheme for image analysis or image enhancement
- G06T2207/20—Special algorithmic details
- G06T2207/20024—Filtering details
- G06T2207/20028—Bilateral filtering
Abstract
The invention discloses a low-quality fingerprint image restoration algorithm and a low-quality fingerprint image restoration device, which belong to the field of image restoration and comprise the following steps: 1) dynamically zooming the low-quality fingerprint image; 2) extracting an image direction field from the zoomed image; 3) carrying out image filtering processing on the original low-quality image; 4) performing image enhancement on the filtered image according to the direction field; 5) thinning the enhanced image; 6) searching for an endpoint in the image; 7) evaluating the continuity of the image lines; 8) comparing the continuity of the image lines with a line continuity threshold, and extracting features of the image meeting the line continuity threshold; and for the image which does not meet the threshold of the continuity of the lines, evaluating the damage degree of the image, repairing the broken lines, and returning to the step 7) after the repairing until the requirement of the threshold of the continuity is met. The method is beneficial to reducing the calculation amount, accelerating the calculation speed, avoiding misconnection in the repair process and reducing the damage to the peripheral line structure.
Description
Technical Field
The invention relates to the technical field of fingerprint image identification and restoration, in particular to a restoration method and a restoration device for a low-quality fingerprint image.
Background
The fingerprint image records the texture characteristics of human fingerprints, and can be used as the unique identification characteristics of each person. Fingerprint image identification is a common biological identification technology, and the identification process is mainly divided into two parts, namely fingerprint feature extraction and fingerprint feature identification.
The fingerprint feature extraction is to simplify fingerprint images, replace curve structures with feature points, reduce calculated amount on the premise of ensuring certain precision and improve operation speed. The fingerprint image collected by the sensor needs to be preprocessed before feature extraction, namely a series of image calculation processes specifically include acquisition of a direction field of the fingerprint image, background separation, equalization, smoothing, binarization and refinement. When extracting the features, firstly searching two types of feature points, filtering partial false feature points, and adding extra information around the feature points to complete the image feature extraction.
The fingerprint feature identification is to judge whether two images belong to the same fingerprint by comparing the features of different fingerprint images. The feature identification process comprises the steps of pre-identification, registration, accurate comparison and the like. The same fingerprint has several fingerprint images, the maximum value produced by the mutual comparison between these fingerprint images and the images produced by other different fingerprints is used as the comparison threshold, the comparison value produced by the comparison with the similar images is divided into two parts by this threshold, and the total number of images with score not less than the threshold is divided by the total number of fingerprint image libraries participating in the comparison process, so as to obtain the identification rate of a set of algorithm corresponding to a set of fingerprint libraries.
In practice, when a fingerprint image is identified, in the process of converting fingerprint biological information into digital information, the fingerprint image is difficult to identify due to the fact that lines of the fingerprint image are broken, deformed and the like caused by the defects of an acquisition instrument and the damage of the fingerprint, and therefore some methods are needed for repairing the defects of the fingerprint image. The present fingerprint image restoration method, for example, the method disclosed in patent No. CN101625724A, includes the steps of performing finite element growth on each binary fingerprint image according to a finite element growth rule with line end points as growth points. The algorithm is only suitable for repairing simple fracture lines which are relatively short in distance and relatively small in direction deviation; for large direction deviation, the calculation amount is large, the efficiency is low, and the processing process is easy to be connected by mistake, so that the surrounding line structure is damaged.
Disclosure of Invention
The invention aims to solve the technical problems that the traditional fingerprint image repairing method is only suitable for repairing simple broken lines which are relatively short in distance and relatively small in direction deviation, and provides a low-quality fingerprint image repairing method and a low-quality fingerprint image repairing device for the broken lines with large direction deviation, which are low in repairing efficiency and easy to be misconnected.
In order to solve the technical problems, the technical scheme provided by the invention is as follows:
the invention relates to a method for restoring a low-quality fingerprint image, which comprises the following steps:
1) carrying out dynamic scaling processing on the original low-quality fingerprint image;
2) extracting an image direction field from the zoomed image;
3) carrying out image filtering processing on the original low-quality fingerprint image;
4) carrying out image enhancement processing on the filtered image according to the direction field;
5) thinning the enhanced image;
6) searching the endpoint in the refined image;
7) evaluating the continuity of the image lines;
8) and setting a line continuity threshold, comparing the continuity of the image lines with the line continuity threshold, repairing broken lines of the image which does not meet the line continuity threshold, and returning to the step 7 after the repairing treatment until the requirement of the continuity threshold is met.
The fingerprint characteristics of the restored image can be further extracted for subsequent image identification, so that the identification rate of the image identification is improved.
Preferably, in step 1), a convolution kernel template based on a gaussian filter is used to perform dynamic scaling on the original low-quality fingerprint image, where the dynamic scaling includes: sampling in the horizontal and vertical directions and reducing the image, sampling and filtering the image, and expanding the gray value of each pixel of the image by a plurality of times upwards to keep the variable of the gray value to generate overflow. The zoomed image is denoised to a certain extent, the characteristics of the original image can be basically reserved, and the calculation of the direction field can be accelerated.
Preferably, after the gray value of each pixel in the image is expanded upwards by a plurality of times, the gray value of each point in the image is calculated to obtain the gray value of each pixel coordinate after expansion, and the calculation formula is as follows:
wherein Z represents the gray value of a certain pixel after expansion, and Z 0 Representing the gray value of a certain pixel point before expansion, K representing the expansion multiple, mod256 representing the remainder of an integer 256, and 256 representing 256 levels of gray value grading.
And selecting a proper expansion multiple, and connecting the breakpoints of the cracks with larger intervals on the premise of keeping the characteristics of the original image in the repairing process of the final image.
And 2) extracting an image direction field from the zoomed image by using a sobel operator in the step 2).
Preferably, in the step 3), dynamic bilateral filtering is adopted to perform filtering processing on the fingerprint image, and a calculation formula is as follows:
wherein the content of the first and second substances,Iis the gray value of the pixel point after filtering,win order to normalize the coefficients of the coefficients,pso as to makeqPoint-centered filtering templatesAt any one point in the above-mentioned (b),I p is a gray-scale value of a p-point,I q is the gray value of the pixel at the center point of the template,
in order to be the spatial domain weight,
in order to be the value range weight,
、
to representpDotx、yThe coordinates of the axes are set to be,
、
representqThe coordinates of the points are determined by the coordinates of the points,
in the form of a spatial domain variance, the variance,
is the value domain variance.
Preferably, before the fingerprint image is filtered by using dynamic bilateral filtering, the two-dimensional bilateral filtering function is split into a filter function in the X-axis direction and a filter function in the Y-axis direction, and the two-dimensional bilateral filtering function is:
after splitting, the filter function in the X-axis direction is:
after splitting, the filter function in the Y-axis direction is:
where σ is the spatial variance, x 0 ，y 0 Is the coordinate of a central pixel, and x and y are the coordinates of any peripheral pixel in the Gaussian template; therefore, during filtering processing, firstly Gaussian filtering is performed in the horizontal direction of the image, then the result is filtered in the vertical direction of the image, and processing efficiency is improved by adopting two one-dimensional convolution operations and a prefabricated Gaussian template; the prefabricated Gaussian template is a fixed one-dimensional Gaussian template generated by adopting a fixed airspace variance value in different fingerprint image libraries.
The two-dimensional bilateral filtering process is divided into two one-dimensional bilateral filtering functions, the results of the X axis and the Y axis of the image are respectively calculated, then matrix operation is carried out, and the operation efficiency can be effectively improved.
Preferably, the step 4) of performing image enhancement processing on the filtered image according to the directional field includes:
4.1) carrying out smooth filtering on the image line direction, and carrying out sharpening filtering on the normal direction of the image line;
4.2) carrying out binarization on the image: approximate Gaussian filtering is adopted in the direction along the lines of the image, and smooth filtering is adopted in the direction of the normal lines of the lines;
4.3) using a gabor filter to perform enhancement processing on the image: and (3) enhancing the image by setting different wavelet scales and directions and combining a direction field.
Preferably, the step 6) searches for end points in the refined image, wherein the end points comprise a close-distance end point of the image, a pseudo-cross point and a turning point of a streak line.
Preferably, when the image continuity is evaluated in step 7), whether an end point of a ruled line belongs to a breakpoint condition is checked in an angle segmentation variable search domain chain mode, that is, whether the end point belongs to a breakpoint is judged by searching whether a potential matching point exists in a space of one end point along the ruled line direction, and the continuity of the ruled line is evaluated according to the number of the breakpoints, where the searching specifically includes:
7.1) selecting an end point, setting the step length and the width of a search domain, advancing a point of the search step length distance along the ridge in the opposite direction of the end point, then sending out from the point, and continuously finding out another point with the same distance along the previous direction;
7.2) connecting the three points in sequence from the end points to form two line segments, and calculating the angle difference of the two line segments as the change value of each angle segment;
7.3) taking the end point as a starting point, making an isosceles triangle along the direction of the end point, wherein the triangle is a single search domain, the step length of the search domain determines the height of the triangle, and the width of the search domain determines the bottom of the triangle;
7.4) in the process of forming the search domain each time, when the extended search domain collides with the nearby ridge, making the base of the triangle in advance from the point where the extended search domain initially meets the nearby ridge, and calculating the distance d between the new base and the original base 1 Thereby calculating the deviation d of the starting point of the next search field 2 Offset d of 2 The calculation formula of (2) is as follows:
wherein, H is the height of the search domain, and L is the length of the bottom of the search domain;
7.5) setting an accumulated value threshold value, recording the accumulated value of the angle change value of each search domain, stopping searching when the accumulated value exceeds the accumulated value threshold value and no matched breakpoint is found, and considering that the two endpoints are not matched and the related line form is complete; and if at least one matched breakpoint is searched, judging that the streak line has a break.
Preferably, the repair processing of the broken striae in step 8) is to perform connection repair according to the matched breakpoint in the process of evaluating the continuity of the striae, and for the search route which is searched within the angle limit range and marked as the breakpoint, that is, along the high connection repair line of each search domain starting from the starting point, the vertex of the search domain and the breakpoint are connected in the last search domain, so as to complete the repair of the striae; and if a plurality of matched target breakpoints exist in the search domain, calculating the vertical distance from the points to the last search domain, and selecting the point with the shortest distance as the most matched point.
The invention also relates to a device for restoring a low-quality fingerprint image, comprising:
1) the dynamic scaling processing module is used for dynamically scaling the original low-quality fingerprint image;
2) the direction field extraction module is used for extracting an image direction field from the zoomed image;
3) the filtering processing module is used for carrying out image filtering processing on the original low-quality fingerprint image;
4) the enhancement processing module is used for carrying out image enhancement processing on the filtered image according to the direction field;
5) the thinning processing module is used for thinning the enhanced image;
6) the searching module is used for searching the endpoints in the thinned image;
7) the evaluation module is used for evaluating the continuity of the image lines;
8) and the repairing module is used for setting a line continuity threshold, comparing the continuity of the image lines with the line continuity threshold, and repairing broken lines of the image which does not meet the line continuity threshold until the requirement of the continuity threshold is met.
Compared with the prior art, the technical scheme provided by the invention has the following beneficial effects:
1. the low-quality fingerprint image restoration method provided by the invention is beneficial to reducing the calculation amount and accelerating the calculation speed by the scheme of dynamically zooming, filtering, enhancing and angle segmenting variable search domain chain on the fingerprint image;
2. the invention adopts the angle segmentation variable search domain chain to carry out continuity analysis, searches the break points and the break points matched with the break points, repairs the mutually matched break points, avoids error connection in the repair process and reduces the damage to the peripheral line structure.
Drawings
FIG. 1 is a flow chart of a method of low quality fingerprint image restoration of the present invention;
FIG. 2 is a flow diagram of breakpoint search, matching, and repair;
FIG. 3.1 is an image scaling pixel expansion diagram for the case where the multiplier K is 16;
FIG. 3.2 is an image scaling pixel expansion diagram for the case where the multiplier K is 8;
FIG. 3.3 is an image scaling pixel expansion diagram for the case where the multiplier K is 4;
FIG. 3.4 is an image scaling pixel expansion diagram for the case where the multiplier K is 2;
FIG. 3.5 is a pixel expansion diagram of image scaling after Gaussian filtering;
FIG. 4.1 is a schematic diagram of an angle segmented variable search domain chain scheme;
FIG. 4.2 is a schematic diagram of a search domain implementing a breakpoint matching scheme;
FIG. 4.3 is a schematic diagram of four search domains completing a breakpoint match matching scheme;
FIG. 4.4 is a schematic diagram of an angle segmented variable search domain chain scheme for search domain correction in combination with surrounding striae;
FIG. 5.1 is a comparison before and after image restoration;
FIG. 5.2 is a comparison before and after another image restoration;
fig. 6 is a schematic block diagram of the low-quality fingerprint image restoration device according to the present invention.
Detailed Description
For further understanding of the present invention, the present invention will be described in detail with reference to examples, which are provided for illustration of the present invention but are not intended to limit the scope of the present invention.
Example 1
Referring to fig. 1, a method for restoring a low-quality fingerprint image includes the following steps:
1) the size of the original low-quality fingerprint image collected in this embodiment is 360 × 256, the image is dynamically scaled, and the original image is scaled by sampling with a gaussian filter whose side length is 3 at 2 times of the sampling interval: in this embodiment, a convolution kernel template based on a gaussian filter is used to perform dynamic scaling on an original low-quality fingerprint image, where the dynamic scaling includes: sampling in the horizontal and vertical directions, reducing the image, sampling and filtering the image, and upwards expanding the gray value of each pixel of the image by a plurality of times to ensure that the gray value is stored to generate overflow;
after upwards expanding the gray value of each pixel of the image by a plurality of times, calculating the gray value of each point in the image to obtain the gray value of each pixel coordinate after expansion, wherein the calculation formula is as follows:
wherein Z represents the gray value of a certain pixel after expansion, and Z 0 Representing the gray value of a certain pixel point before expansion, K representing the expansion multiple, mod256 representing the remainder of an integer 256, and 256 representing 256 levels of gray value grading.
For example, the gray scale value region of 0 to 84 is set as a dark region, 86 to 169 are set as a middle region, 170 to 255 are set as a light region, K =2, and Z is set 0 In the gray distribution values of 0-255, the gray values closer to the two ends have smaller change degree after expansion; the pixels distributed in the middle area are converted into darker pixels due to variable overflow; the pixels in the dark areas become brighter. On one hand, the area where the original lines are located shrinks towards the ridge direction of the lines, a part of pixel points located at the edges of the lines become lighter, and the pixels with enough dark gray levels originally keep dark colors. Meanwhile, pixels originally at the middle gray level near the streak line are shifted toward a dark color, and dispersed black dots are formed in a light color area. In general, the dark pixels at the edge of the lines are diffused, and some black dispersion points appear in the transition areas of different lines. The closer the diffusion effect is to the lines, the stronger the diffusion effect is, the gray in the transition areaThe closer the value is to the average value of the gray scale, the more likely the black dots are generated. These diffused or generated black dots are connected to each other as short lines of several pixels long, filling in between the line break points and the line horizontal spaces. The short lines between the break points of the lines weaken the gradient value of the field in the region direction, so that the two break points with the same direction are associated, the trend of mutual connection in the normal direction of the lines at the crack is counteracted, the change of the direction field is smoother, and the break points are favorably connected into lines during image enhancement and binaryzation. Meanwhile, because the normal direction discrimination at the continuous part of the lines is large enough, the weights of the scattered points and the short lines are small when the direction field of the area is calculated, and almost no influence is caused.
Therefore, most scattered points and short lines generated among the fingerprint lines are beneficial to smoothing the direction field between the break points, so that the break points are connected into lines in the subsequent image enhancement processing, and the calculation of the direction field of other areas cannot be interfered. When the expansion multiple K is increased, the pixel point change distribution of each region in the original thumbnail is more dispersed. For example, when K = 3, the pixel set with the gray values distributed in 0 ~ 85 is expanded and separated into three levels with sequentially increasing gray levels. When K = 6, the set is separated into two sets of hierarchical blocks with sequentially increasing gray levels. The image has the advantages that the streak line diffusion effect is stronger, and the image generates more noise points.
For fig. 3.1-3.5, the picture quality is sequentially reduced and the noise is increased as a whole. From the detail, the original black line part pixel is diffused to the periphery and becomes thin gradually; the white gap lines have gray or black scattered points, and originally broken intervals are filled; the spatial regions closer to the streak line are assigned more scattered points, the more drastic the change, and the spatial gray values farther from the streak line are relatively changed less or more smoothly. When the black pixels of the lines are properly diffused, the lines themselves become relatively thin, and the color of the light-colored spaces is locally enhanced. However, when the black pixel is diffused to a large extent, the streak lines gradually mingle with the gaps, and the overall image quality is therefore degraded.
The gray value expansion is the dynamic embodiment in the dynamic scaling process. After the Gaussian filtering operation is completed, any point in the newly formed scaled image is the weighted value of the pixel point of the original image in the Gaussian template range.
FIG. 3.5 is the image after Gaussian filtering, the shapes of the gaps between dark lines and light lines are clear, and the gaps between cracks and normal lines are clearly distinguished;
fig. 3.4 shows the case that the multiplier K is 2, that is, after the gray value of each pixel point is expanded by 2 times, 256 is left, and the streak line becomes slightly thinner, but the original form is basically maintained; the gaps and the cracks are filled with a plurality of noise points;
FIG. 3.3 shows the case where the multiplier K is 4, the lines are further thinned, the gray scale of the noise points between the gaps is enhanced, and the noise points generated by the fingerprint region tend to spread out of the fingerprint image region;
FIG. 3.2 shows that when the multiplier K is 8, the original shapes of the lines and spaces begin to change and have a tendency of merging with each other; image noise points are further increased, and obvious light color noise points also appear in the edge area inside the grain line;
fig. 3.1 shows that with a multiplier K of 16, both the fingerprint image area and the background appear a lot of noise, and the ridge shape is hardly recognizable.
Therefore, the zoomed image is added with certain controllable noise in the filtered image, so that the originally broken gap is connected in a scattered manner, the influence on the normal fingerprint ridge gap and the influence outside the fingerprint area are relatively low, and the probability of connection of the broken ridge during image enhancement is improved on the premise of basically keeping the characteristics of the original image. Meanwhile, the size of the zoomed image is only a fraction of that of the original image, and the calculation of the direction field is accelerated. From the comparison of the five figures, K =2 or 4 is better for the subsequent repair, and is a suitable expansion factor, while K =8 or 16 is worse.
2) And extracting an image direction field from the zoomed image by using a sobel operator.
3) The dynamic bilateral filtering is adopted to carry out image filtering processing on the original low-quality fingerprint image, and the calculation formula is as follows:
wherein the content of the first and second substances,Iis the gray value of the pixel point after filtering,win order to normalize the coefficients of the coefficients,pso as to makeqPoint-centered filtering templatesAt any one point in the above-mentioned (b),I p is a gray-scale value of a p-point,I q is the gray value of the pixel at the center point of the template,
a spatial domain weight, a value domain weight,
、
to representpDotx、yThe coordinates of the axes are set to be,
、
representqThe coordinates of the points are determined by the coordinates of the points,
is a variance in the spatial domain and is,
is the value domain variance.
In this embodiment, the two-dimensional bilateral filter function is:
before the fingerprint image is filtered by adopting dynamic bilateral filtering, the two-dimensional bilateral filtering function can be divided into a filter function in the X-axis direction and a filter function in the Y-axis direction,
after splitting, the filter function in the X-axis direction is:
after splitting, the filter function in the Y-axis direction is:
where σ is the spatial variance, x 0 ，y 0 Is the coordinate of the central pixel, and x and y are the coordinates of any surrounding pixel in the Gaussian template.
The Gaussian template can adopt fixed airspace variance values for different fingerprint image libraries, generates a fixed one-dimensional Gaussian template, and adopts fixed point number calculation, so that the operation efficiency is effectively improved on the premise of ensuring certain precision, for example, the identification rate of a certain fingerprint image library takes sigma in the airspace variance 0 Time reaches maximum, at this time, one S × S normalized gaussian template is:
the template may be decomposed into the form of the product of a column vector and a row vector, such as:
wherein for any i e [1, S]All have | x i | = | y i Therefore, the normalized one-dimensional gaussian template with the fixed length of S is firstly subjected to gaussian filtering in the horizontal direction of the image, and then the result is subjected to filtering in the vertical direction of the image, so that the fact that the operation speed can be effectively accelerated by adopting two one-dimensional convolution operations and the prefabricated gaussian template, namely, the fast bilateral filtering can be realized.
Let the gray value of a certain central pixel be Z 0 The gray value of a certain peripheral pixel is Z ij Where i represents the X-axis offset of the surrounding pixels relative to the center pixel and j is the Y-axis offset. The absolute value of the difference between the gray values of the two pixels
Sum and difference value interval
Comprises the following steps:
the value domain weights of the bilateral filter within the convolution template are:
therefore, the larger the value of the value domain variance is, the smaller the weight ratio of the maximum difference value and the minimum difference value in the value domain is, and the lower the selection effect on the space domain is. If the value range variance approaches infinity, the bilateral filter only keeps the Gaussian filtering effect; if the value is close to infinitesimal, the Gaussian filtering effect is completely eliminated, and the bilateral filtering result is basically equal to the original image. For example, if
The value of the gray-level difference is the maximum valueThe domain weight is exp (-1) = 0.368, the weight is exp (0) = 1 when the weight is the minimum value, and the effect of the Gaussian filter is reduced by about 65% due to the lowest weight of the range; when in use
And then, the value domain weight of the maximum gray value difference value is exp (-1) = 0.018, and the comprehensive weight of the corresponding pixel point is greatly reduced.
The bilateral filter in the invention uses a processing method for color noise in the process of processing a color image, namely, a RGB color model is transferred to a YUV model by contrast, so that the lightness and the color attribute of the color image are separated, and a method of different bilateral filter airspace and value domain variances is adopted for two channels, so that the process of effectively reducing the color noise is achieved. Under the condition that the position of the estimated cord fracture point is low, namely the continuity of the cord is low, the weight of a value range in the bilateral filter is reduced, the Gaussian filter plays a main role, the part of the original image which is judged as the edge by mistake is treated as a part of the continuous cord, the higher probability of being connected in the subsequent image enhancement process is ensured, and the purpose of repairing the cord in the filtering process is achieved.
In summary, the two-dimensional bilateral filtering process is divided into two one-dimensional bilateral filtering functions, and the results of the X and Y axes of the image are calculated respectively, and then matrix operation is performed, so that the operation efficiency can be effectively improved.
4) The image enhancement processing is carried out on the filtered image according to the direction field, and the specific steps are as follows:
4.1) setting a gabor wavelet transform with 12 directions and 3 dimensions, adopting smooth filtering with a template of [1, 1, 1, 1, 1, 1, 1] in the horizontal direction, and adopting sharpening filtering with a template of [ -3, -1, 3, 9, 3, -1, -3] in the vertical direction, further performing smooth filtering on the image line direction, and performing sharpening filtering on the image line normal direction to realize image enhancement;
4.2) carrying out binarization on the image: approximate Gaussian filtering is performed by adopting a template of [1, 2, 4, 5, 4, 2, 1] along the streak line direction of the image, and smooth filtering is performed by adopting a template of [1, 1, 1, 1, 1, 1, 1, 1] along the normal line direction of the streak line;
4.3) using a gabor filter to perform enhancement processing on the image: and (3) enhancing the image by setting different wavelet scales and directions and combining a direction field.
And by setting different wavelet scales and directions and combining a direction field, better enhancement processing can be realized on the image.
5) Thinning the enhanced image, wherein thinning in the horizontal direction is performed first, and thinning in the vertical direction is performed.
6) Searching for end points in the refined image, as shown in fig. 2, the end points include a short-distance end point, a pseudo-cross point and a turning point of the image, and this embodiment considers not only all normal short-distance end points but also special-form cross points (pseudo-cross points) and turning points of some lines, because: breakpoints of crack edges possibly generated due to the calculation deviation of the direction field in the low-quality fingerprint image are easily connected into a line, so that the original points which are the breakpoints are changed into cross points; different from the common fork points, the included angle of the two lines of the crack edge fork points is larger, and the pseudo fork points can be identified by detecting the direction field of the position near each fork point and are used as end points for calculation; in addition, some lines with drastic local direction change exist, which is characterized in that the direction of the lines is approximately vertically changed within a short distance of a few pixels. These turning regions may be originally a breakpoint, cannot be determined by simply determining the sum of absolute differences between the end point and the point near the crossing point, and should be treated as the end point as well.
7) The method comprises the steps of starting from each breakpoint by adopting a search domain which is approximately in a regular triangle and has a search width of 10, searching along the direction of a line where the breakpoint is located by using an angle subsection variable search domain chain mode, determining the change angle value of each section of the search domain according to the angle difference of the starting line segment of the line where the breakpoint is located, checking whether the endpoint of the line belongs to the breakpoint condition, and judging whether the endpoint belongs to the breakpoint by searching whether a potential matching point exists in the space of one endpoint along the direction of the line, namely in the search process, when at least one other breakpoint is detected within the change angle of more than 90 degrees, considering that the line has the fracture condition, and evaluating the continuity of the line according to the number of the breakpoints.
The present embodiment sets the direction field of the end point existing position to be flat and gentle, where the flat is generally far from the singular point and the gentle is generally close to the singular point. The distance between the two end points is divided into a short distance and a long distance, so that four combination modes exist: short-distance breakage at a gentle change, long-distance breakage at a gentle change, short-distance breakage at a violent change, and long-distance breakage at a violent change. Because the straight-line distances of the breakpoints near the singular points are all close, the long-distance fracture condition at the violent change position can be practically not considered.
If an end point is a break point of a ruled line, at least one end point should exist in the space along the direction of the ruled line, so that under certain conditions, two end points can be connected into an approximate curve consisting of one or more line segments. The space along the direction of the thread line is covered by a search chain consisting of a plurality of sections of search domains, and the search step length and the search width of each search domain are set to control the size of each search domain.
Referring to fig. 4.1-4.4, a detailed description of the variable search domain chain scheme for angular segmentation from top to bottom is provided:
as shown in fig. 4.1, starting from the position determined as the break point along the opposite direction of the cord where the point is located (taking the counterclockwise direction as positive), the first point is found along the length of the forward D1 of the cord, and the second point is found from the length of the forward D1 of the point. Taking the second point as a starting point, and making a ray passing through the first point; and then taking the first point as a starting point, and making a ray passing through the breakpoint. The acute included angle θ formed by the two rays is taken as the angular offset that each time a new search field is formed. Starting from the break point along the line direction of the point, the break point is taken as a vertex, the direction field direction of the point is taken as a height with the length of D1, the normal direction of the other end of the height is taken as a line segment with the length of D2, and the midpoint of the line segment is the intersection point of the line segment and the height. And connecting the break point and the two ends of the line segment with the length of D2 to form an isosceles triangle, wherein the isosceles triangle is the first search domain. If no matched breakpoint is found in the search field, starting from the middle point of the bottom edge of the search field, shifting by an angle theta along the counterclockwise direction, repeating the process of establishing the search field, and recording the accumulated angle shift.
And before the accumulated offset angle does not exceed a certain threshold, finding at least one matched breakpoint in the search domain chain, and considering that the lines where the two breakpoints are located are broken and can be connected into one line. If the accumulated offset angle exceeds the set threshold, the breakpoint is considered to be a common endpoint, and the shape of the line where the breakpoint is located is complete.
Fig. 4.2 shows an ideal short-distance search process, i.e. matching of breakpoints is completed by establishing only one search field, which is a summary of the search method.
Fig. 4.3 shows the parameter characteristics of the search field, i.e. the high of the search field determines the search step size and the bottom of the search field determines the search range. The shorter the step length is, the finer the search is, and the longer the time consumption is; the wider the range, the more likely a matching point is to be searched, and the search time may be increased and the erroneous judgment may be increased.
Fig. 4.4 shows an ideal search chain forming process, in which point P1 is the starting point of the search and point P2 is the ending point of the search, and the search process is referred to above.
In the actual search process, in addition to the position and angle of the ridge itself, the characteristics of the surrounding ridge must be considered. When the form of the adjacent lines of the fracture line is intact, the breakpoint searching and matching process of the fracture line can play a role in guiding and limiting, so that the searching result is more reliable. Fig (4.5) shows a more complex search process: starting from the right point of the figure, a plurality of search domains are formed in sequence. The search field represented by the dashed line is the original field and the search field represented by the solid line is the modified field. The purpose of correction is to limit and guide the matching point search space of the break points on the lines by using the forms of the adjacent lines.
The correction method is divided into two types: area correction and translation correction. The priority of area correction is higher than that of translation correction, but continuous area correction can not occur. When any one side of the newly established search domain contacts with the two side lines, the intersection point of the side which is contacted first is taken as the end point of the correction bottom edge, and the search domain is reestablished. And reducing the area of the search domain to avoid collision with adjacent lines, namely area correction. The area correction only affects the current search field, and the high and bottom side lengths of the new search field are still determined by the already set values of D1 and D2. If the area of the previous search domain is corrected in the process of establishing the new search domain, and any side of the new search domain is still in contact with the two side lines after the new search domain is established, the search domain is translated for a short distance in the opposite direction of the side in which the search domain is in contact with the first side by taking the established vertex as a starting point. And if the search domain does not collide with the adjacent lines after translation, searching in the search domain according to a normal flow. And if the collision still occurs after the translation, performing area correction.
The specific process illustrated in fig (4.5) is as follows: starting from the right point, the original domain of the first search domain collides with the lower adjacent ridge, and the area correction is performed, which is embodied as a distance d1 being shortened in the high direction. The second search field collides with the lower adjacent stripe line, and because the area correction is already carried out on the previous search field, the search field firstly carries out translation correction, and no collision occurs after correction. The establishment of subsequent search domains repeats the above process. The purpose of using the translation correction is to avoid that in the case of an inappropriate offset angle, a new search field may become smaller and smaller due to the area correction, and the meaning of the search is lost.
Referring to fig. 1, 2 and 4, the searching steps in the angle segmentation variable search domain chain mode are as follows:
7.1) selecting an end point, setting the step length and the width of a search domain, advancing a point of the search step length distance along the ridge in the opposite direction of the end point, then sending out from the point, and continuously finding out another point with the same distance along the previous direction;
7.2) connecting the three points in sequence from the end points to form two line segments, and calculating the angle difference of the two line segments as the change value of each angle segment;
7.3) taking the end point as a starting point, making an isosceles triangle along the direction of the end point, wherein the triangle is a single search domain, the step length of the search domain determines the height of the triangle, and the width of the search domain determines the bottom of the triangle;
7.4) in the process of forming the search domain each time, when the extended search domain collides with the nearby ridge, making the base of the triangle in advance from the point where the extended search domain initially meets the nearby ridge, and calculating the distance d between the new base and the original base 1 Thereby calculating the deviation d of the starting point of the next search field 2 Offset d of 2 The calculation formula of (2) is as follows:
wherein, H is the height of the search domain, and L is the length of the bottom of the search domain;
the purpose of using the offset is to ensure that the search domain is always kept in a reasonable detection range and cannot exceed a space limited by the attached adjacent lines, and a potential matching breakpoint is effectively found, wherein the offset direction is determined by a direction field of a starting breakpoint and one side of a first contact edge of two waists of the search domain, for example, when the direction field of the starting breakpoint is 0-90 degrees, if the right waist of the search domain contacts the adjacent line first, the offset direction of the next time should be leftward;
7.5) for the fracture situation at the severe change, the search step length is reduced, the search range is enlarged, on the end edge of each search domain, the intersection point of the top and the bottom is taken as the starting point of the next search domain, the angle difference of two line segments of the cord where the original end point is located is taken as the angle variable, the shape is reset and different search domains are oriented, and thus the adjustment of the angle of the search domain is completed, therefore, the embodiment sets the threshold value of the accumulated value, records the accumulated value of the angle change value of each search domain, and stops searching when the accumulated value exceeds the threshold value of the accumulated value and no matched breakpoint is found, the two end points are considered to be not matched, and the related cord is complete in form; and if at least one matched breakpoint is searched, judging that the streak line has a break.
8) Setting a line continuity threshold, comparing the continuity of the image lines with the line continuity threshold, and extracting features for subsequent image identification of the image meeting the line continuity threshold; for the image which does not meet the threshold of the continuity of the lines, repairing the broken lines, and returning to the step 7) after the repairing until the requirement of the threshold of the continuity is met;
the broken lines are repaired according to the connection and repair of the matched breakpoints in the process of evaluating the continuity of the lines, and the search routes marked as the breakpoints are searched in the angle limit range, namely, the top points and the breakpoints of the search domains are connected in the last search domain along the high connection repair lines of each search domain starting from the starting point, so that the repair of the lines is completed; and if a plurality of matched target breakpoints exist in the search domain, calculating the vertical distance from the points to the last search domain, and selecting the point with the shortest distance as the most matched point.
Fig. 5.1 and 5.2 show comparison examples before and after restoration of two sets of fingerprint images. In the whole view, each group of images sequentially shows the original image of the fingerprint image, the repairing effect of the general method and the repairing effect of the patent method from left to right.
From the original image, the fracture lines of the upper image are narrower than those of the lower image, and the overall line shape is complete. However, from the repair situation of the general method, both of them have local disorder of the line shape, which means that the quality of the fingerprint image cannot be judged only by naked eyes, and the quality of the image should be measured by the standard of the processing difficulty of the method.
From the two original images, it can be seen that the broken lines and local breakage of the fingerprint images result in that the true features of the fingerprint cannot be restored when processed by the general method, and meanwhile, some false features which the original images do not have are generated. The processing result of the general method shows that the processing result is better in the clear and complete part of the fingerprint original image lines; however, in the area where the image lines are broken and damaged, the general method processes the original broken points into cross points, connects the points into new lines, and changes the trend of the adjacent lines. In addition, the damaged area of the image also generates some mixed ring structures and directional fields, which indicates that the original characteristics of the fingerprint cannot be well recovered when the general processing method is faced with cracks and damages.
The rightmost image of each set of images shows the results of the method proposed by this patent. It is clear that most of the broken lines are repaired, the overall direction field change also approaches normal, and at the same time, too many false features are not generated. Of course, the original image features are still not completely restored or restored in the processing result graph of the method provided by the patent, the main reason is that different search domain feature values have different processing capabilities for different fracture structures, but if the search domain feature values are set according to specific structures, the complexity of calculation is increased, and therefore, the accuracy and the time cost of the method need to be considered at the same time.
The following table shows the statistical results of the processing of two sets of fingerprint image libraries by a general method and the method proposed in this patent:
description of the drawings:
1. the two fingerprint libraries are all different fingerprint images which are acquired by the capacitance type fingerprint acquisition equipment and have the size of 360 multiplied by 256 pixels. Each identical fingerprint is repeatedly collected for 10 times, so that each picture of the identical fingerprint is slightly different;
2. the general method refers to a method similar to the method described in fig. 1 except for the gray scale expansion, dynamic bilateral filtering, streak line continuity evaluation, streak line restoration method and each filtering template value at the image filtering and enhancing stage proposed in the patent method;
3. the heterogeneous maximum value refers to the maximum similarity score obtained when two different fingerprints are compared;
4. the zero false recognition rate means that the maximum value of different types is used as a threshold, and when the similarity score of the same fingerprint image is smaller than the threshold, the recognition is considered to be failed. The ratio of the difference between the total number of times of the similar image comparison and the total number of the recognition failure images in the total number of times of the similar image comparison is zero false recognition rate.
Example 2
Referring to fig. 6, the present embodiment relates to a low-quality fingerprint image restoration apparatus, including:
1) the dynamic scaling processing module is used for dynamically scaling the original low-quality fingerprint image; the dynamic scaling processing module is used for realizing the function of step 1) in the embodiment 1.
2) The direction field extraction module is used for extracting an image direction field from the zoomed image; the direction field extraction module is used for realizing the function of step 2) in the embodiment 1.
3) The filtering processing module is used for carrying out image filtering processing on the original low-quality fingerprint image; the filtering processing module is used for realizing the function of step 3) in the embodiment 1.
4) The enhancement processing module is used for carrying out image enhancement processing on the filtered image according to the direction field; the enhancement processing module is used for realizing the function of the step 4) of the embodiment 1.
5) The thinning processing module is used for thinning the enhanced image; the refinement processing module is used for realizing the function of step 5) in the embodiment 1.
6) The searching module is used for searching the endpoints in the thinned image; the searching module is used for realizing the function of the step 6) of the embodiment 1.
7) The evaluation module is used for evaluating the continuity of the image lines; the evaluation module is used for realizing the function of step 7) of the embodiment 1.
8) The restoration module is used for setting a line continuity threshold, comparing the continuity of the image lines with the line continuity threshold, and extracting features for subsequent image identification of the image meeting the line continuity threshold; and for the image which does not meet the threshold of the continuity of the lines, repairing the broken lines until the requirement of the threshold of the continuity is met. The repair module is used for realizing the function of step 8) in the embodiment 1.
Obviously, the repair apparatus of the present embodiment can be an execution subject of the repair method of embodiment 1 described above, and thus can realize the functions realized by the repair method. Since the principle is the same, the detailed description is omitted here.
The present invention has been described in detail with reference to the embodiments, but the description is only for the preferred embodiments of the present invention and should not be construed as limiting the scope of the present invention. All equivalent changes and modifications made within the scope of the present invention shall fall within the scope of the present invention.
Claims (10)
1. A method for restoring a low-quality fingerprint image, comprising: which comprises the following steps:
1) carrying out dynamic scaling processing on the original low-quality fingerprint image;
2) extracting an image direction field from the zoomed image;
3) carrying out image filtering processing on the original low-quality fingerprint image;
4) carrying out image enhancement processing on the filtered image according to the direction field;
5) thinning the enhanced image;
6) searching the endpoint in the refined image;
7) evaluating the continuity of the image lines;
8) and setting a line continuity threshold, comparing the continuity of the image lines with the line continuity threshold, repairing broken lines of the image which does not meet the line continuity threshold, and returning to the step 7 after the repairing treatment until the requirement of the continuity threshold is met.
2. A method of inpainting a low-quality fingerprint image as recited in claim 1, wherein: in the step 1), a convolution kernel template based on a Gaussian filter is adopted to perform dynamic scaling processing on the original low-quality fingerprint image, wherein the dynamic scaling processing comprises the following steps: sampling in the horizontal and vertical directions and reducing the image, sampling and filtering the image, and expanding the gray value of each pixel of the image by a plurality of times upwards to keep the variable of the gray value to generate overflow.
3. The method for inpainting a low-quality fingerprint image according to claim 2, wherein: after upwards expanding the gray value of each pixel of the image by a plurality of times, calculating the gray value of each point in the image to obtain the gray value of each pixel coordinate after expansion, wherein the calculation formula is as follows:
wherein Z represents the gray value of a certain pixel after expansion, and Z 0 Representing the gray value of a certain pixel point before expansion, K representing the expansion multiple, mod256 representing the remainder of an integer 256, and 256 representing 256 levels of gray value grading.
4. A method of inpainting a low-quality fingerprint image as recited in claim 1, wherein: in the step 3), dynamic bilateral filtering is adopted to perform filtering processing on the fingerprint image, and the calculation formula is as follows:
wherein the content of the first and second substances,Iis the gray value of the pixel point after filtering,win order to normalize the coefficients of the coefficients,pso as to makeqPoint-centered filtering templatesAt any one point in the above-mentioned (b),I p is a gray-scale value of a p-point,I q is the gray value of the pixel at the center point of the template,
a spatial domain weight, a value domain weight,
、
to representpDotx、yThe coordinates of the axes are set to be,
、
to representqThe coordinates of the points are determined by the coordinates of the points,
in the form of a spatial domain variance, the variance,
is the value domain variance.
5. The method for inpainting a low-quality fingerprint image according to claim 4, wherein: when the dynamic bilateral filtering is adopted to carry out filtering processing on the fingerprint image, a two-dimensional bilateral filtering function is split into a filter function in the X-axis direction and a filter function in the Y-axis direction, and the two-dimensional bilateral filtering function is as follows:
after splitting, the filter function in the X-axis direction is:
after splitting, the filter function in the Y-axis direction is:
where σ is the spatial variance, x 0 ，y 0 Is the coordinate of a central pixel, and x and y are the coordinates of any peripheral pixel in the Gaussian template; therefore, during filtering processing, firstly Gaussian filtering is performed in the horizontal direction of the image, then the result is filtered in the vertical direction of the image, and processing efficiency is improved by adopting two one-dimensional convolution operations and a prefabricated Gaussian template; the prefabricated Gaussian template is a fixed one-dimensional Gaussian template generated by adopting a fixed airspace variance value in different fingerprint image libraries.
6. A method of inpainting a low-quality fingerprint image as recited in claim 1, wherein: the step 4) of performing image enhancement processing on the filtered image according to the direction field comprises the following specific steps:
4.1) carrying out smooth filtering on the image line direction, and carrying out sharpening filtering on the normal direction of the image line;
4.2) carrying out binarization on the image: approximate Gaussian filtering is adopted in the direction along the lines of the image, and smooth filtering is adopted in the direction of the normal lines of the lines;
4.3) using a gabor filter to perform enhancement processing on the image: and (3) enhancing the image by setting different wavelet scales and directions and combining a direction field.
7. A method of inpainting a low-quality fingerprint image as recited in claim 1, wherein: and 6) searching end points in the refined image in the step 6), wherein the end points comprise a short-distance end point, a pseudo-cross point and a turning point of a streak line of the image.
8. The method for inpainting a low-quality fingerprint image of claim 7, wherein: when the image continuity is evaluated in the step 7), whether the end point of the ruled line belongs to the breakpoint condition is checked by adopting an angle segmentation variable search domain chain mode, that is, whether the end point belongs to the breakpoint is judged by searching whether a potential matching point exists in the space of one end point along the ruled line direction, and the continuity of the ruled line is evaluated according to the number of the breakpoints, wherein the searching specifically comprises the following steps:
7.1) selecting an end point, setting the step length and the width of a search domain, advancing a point of the search step length distance along the ridge in the opposite direction of the end point, then sending out from the point, and continuously finding out another point with the same distance along the previous direction;
7.2) connecting the three points in sequence from the end points to form two line segments, and calculating the angle difference of the two line segments as the change value of each angle segment;
7.3) taking the end point as a starting point, making an isosceles triangle along the direction of the end point, wherein the triangle is a single search domain, the step length of the search domain determines the height of the triangle, and the width of the search domain determines the bottom of the triangle;
7.4) in the process of forming the search domain each time, when the extended search domain collides with the nearby ridge, making the base of the triangle in advance from the point where the extended search domain initially meets the nearby ridge, and calculating the distance d between the new base and the original base 1 Thereby calculating the deviation d of the starting point of the next search field 2 Offset d of 2 The calculation formula of (2) is as follows:
wherein, H is the height of the search domain, and L is the length of the bottom of the search domain;
7.5) setting an accumulated value threshold value, recording the accumulated value of the angle change value of each search domain, stopping searching when the accumulated value exceeds the accumulated value threshold value and no matched breakpoint is found, and considering that the two endpoints are not matched and the related line form is complete; and if at least one matched breakpoint is searched, judging that the streak line has a break.
9. A method of inpainting a low-quality fingerprint image as recited in claim 8, wherein: the broken striae in the step 8) is repaired by connection according to the matched breakpoints in the process of evaluating the continuity of the striae, and the search route marked as the breakpoints is searched in the angle limit range, namely, the top point and the breakpoints of the search domain are connected in the last search domain along the high-connection repair line of each search domain starting from the starting point, so that the repair of the striae is completed; and if a plurality of matched target breakpoints exist in the search domain, calculating the vertical distance from the points to the last search domain, and selecting the point with the shortest distance as the most matched point.
10. A device for restoring a low-quality fingerprint image, comprising: it includes:
1) the dynamic scaling processing module is used for dynamically scaling the original low-quality fingerprint image;
2) the direction field extraction module is used for extracting an image direction field from the zoomed image;
3) the filtering processing module is used for carrying out image filtering processing on the original low-quality fingerprint image;
4) the enhancement processing module is used for carrying out image enhancement processing on the filtered image according to the direction field;
5) the thinning processing module is used for thinning the enhanced image;
6) the searching module is used for searching the endpoints in the thinned image;
7) the evaluation module is used for evaluating the continuity of the image lines;
8) the restoration module is used for setting a line continuity threshold, comparing the continuity of the image lines with the line continuity threshold and extracting the characteristics of the image meeting the line continuity threshold; and for the image which does not meet the threshold of the continuity of the lines, repairing the broken lines until the requirement of the threshold of the continuity is met.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202210785657.8A CN114863492B (en) | 2022-07-06 | 2022-07-06 | Method and device for repairing low-quality fingerprint image |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202210785657.8A CN114863492B (en) | 2022-07-06 | 2022-07-06 | Method and device for repairing low-quality fingerprint image |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN114863492A true CN114863492A (en) | 2022-08-05 |
| CN114863492B CN114863492B (en) | 2022-09-13 |
Family
ID=82625825
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202210785657.8A Active CN114863492B (en) | 2022-07-06 | 2022-07-06 | Method and device for repairing low-quality fingerprint image |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN114863492B (en) |
Cited By (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN116823679A (en) * | 2023-08-30 | 2023-09-29 | 山东龙腾控股有限公司 | Full-automatic fingerprint lock fingerprint image enhancement method based on artificial intelligence |
Citations (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US20100303310A1 (en) * | 2009-05-28 | 2010-12-02 | Li-Kuo Chiu | High-performance fingerprint image-processing method |
| CN102368333A (en) * | 2011-09-07 | 2012-03-07 | 常州蓝城信息科技有限公司 | Fingerprint enhancement method based on time domain and frequency domain filtering |
| CN109101921A (en) * | 2018-08-10 | 2018-12-28 | 安徽师范大学 | A kind of Fingerprint diretion extracting method of bonding position diffusion |
| CN112115848A (en) * | 2020-09-16 | 2020-12-22 | 哈尔滨理工大学 | Incomplete area processing method of low-quality fingerprint image |
-
2022
- 2022-07-06 CN CN202210785657.8A patent/CN114863492B/en active Active
Patent Citations (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US20100303310A1 (en) * | 2009-05-28 | 2010-12-02 | Li-Kuo Chiu | High-performance fingerprint image-processing method |
| CN102368333A (en) * | 2011-09-07 | 2012-03-07 | 常州蓝城信息科技有限公司 | Fingerprint enhancement method based on time domain and frequency domain filtering |
| CN109101921A (en) * | 2018-08-10 | 2018-12-28 | 安徽师范大学 | A kind of Fingerprint diretion extracting method of bonding position diffusion |
| CN112115848A (en) * | 2020-09-16 | 2020-12-22 | 哈尔滨理工大学 | Incomplete area processing method of low-quality fingerprint image |
Cited By (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN116823679A (en) * | 2023-08-30 | 2023-09-29 | 山东龙腾控股有限公司 | Full-automatic fingerprint lock fingerprint image enhancement method based on artificial intelligence |
| CN116823679B (en) * | 2023-08-30 | 2023-12-05 | 山东龙腾控股有限公司 | Full-automatic fingerprint lock fingerprint image enhancement method based on artificial intelligence |
Also Published As
| Publication number | Publication date |
|---|---|
| CN114863492B (en) | 2022-09-13 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| US11922615B2 (en) | Information processing device, information processing method, and storage medium | |
| US11238274B2 (en) | Image feature extraction method for person re-identification | |
| CN110678901B (en) | Information processing apparatus, information processing method, and computer-readable storage medium | |
| CN107679520B (en) | Lane line visual detection method suitable for complex conditions | |
| CN113658132B (en) | Computer vision-based structural part weld joint detection method | |
| CN104751142B (en) | A kind of natural scene Method for text detection based on stroke feature | |
| WO2018145470A1 (en) | Image detection method and device | |
| CN115345885A (en) | Method for detecting appearance quality of metal fitness equipment | |
| CN109540925B (en) | Complex ceramic tile surface defect detection method based on difference method and local variance measurement operator | |
| CN109241973B (en) | Full-automatic soft segmentation method for characters under texture background | |
| CN114820625B (en) | Automobile top block defect detection method | |
| CN111354047B (en) | Computer vision-based camera module positioning method and system | |
| CN112991283A (en) | Flexible IC substrate line width detection method based on super-pixels, medium and equipment | |
| CN111402316A (en) | Method for rapidly detecting ellipse in image based on false link removal | |
| CN110348307B (en) | Path edge identification method and system for crane metal structure climbing robot | |
| CN114863492B (en) | Method and device for repairing low-quality fingerprint image | |
| JP4062987B2 (en) | Image area dividing method, image area dividing apparatus, and image area dividing program | |
| CN112508913A (en) | Cable section edge detection method based on image detection | |
| CN110427979B (en) | Road water pit identification method based on K-Means clustering algorithm | |
| CN114863493B (en) | Detection method and detection device for low-quality fingerprint image and non-fingerprint image | |
| CN108205641B (en) | Gesture image processing method and device | |
| CN112597868A (en) | Test paper identification and correction method based on no positioning point | |
| CN113516121A (en) | Multi-feature fusion non-motor vehicle license plate region positioning method | |
| JP2871590B2 (en) | Image extraction method | |
| JPH0468481A (en) | Character segmenting device |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| GR01 | Patent grant | ||
| GR01 | Patent grant |