CN115797186A - Image restoration method and device, electronic equipment and storage medium - Google Patents
Image restoration method and device, electronic equipment and storage medium Download PDFInfo
- Publication number
- CN115797186A CN115797186A CN202111015407.8A CN202111015407A CN115797186A CN 115797186 A CN115797186 A CN 115797186A CN 202111015407 A CN202111015407 A CN 202111015407A CN 115797186 A CN115797186 A CN 115797186A
- Authority
- CN
- China
- Prior art keywords
- block
- repaired
- pixel point
- image
- current
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Pending
Links
Images
Abstract
The invention relates to an image restoration method, an image restoration device, an electronic device and a storage medium, wherein the method comprises the following steps: determining the boundary of a current region to be repaired of a first image; determining a current sample block size; generating a first block to be repaired taking a first pixel point as a center and a neighborhood block set adjacent to the first block to be repaired according to the size of a current sample block; wherein, the first pixel point is one of the pixel point sets on the boundary; acquiring a correlation coefficient between a first block to be repaired and a neighborhood block set; determining the priority of the first pixel point according to the correlation coefficient, the confidence coefficient of the first pixel point and the data item of the first pixel point; and determining the current pixel point to be repaired and a target matching block corresponding to the current pixel point to be repaired according to the priority of each pixel point in the pixel point set on the boundary, and repairing the block to be repaired corresponding to the current pixel point to be repaired according to the target matching block. The pixel point priority result obtained by the method is more comprehensive and accurate, and the image restoration quality and efficiency are comprehensively improved.
Description
Technical Field
The present invention relates to the field of image processing technologies, and in particular, to an image restoration method and apparatus, an electronic device, and a storage medium.
Background
In clinical medicine at present, diagnosis and treatment are often assisted by taking medical images, for example, CT images are taken for pathological analysis, and marker balls (Mark balls) which can be developed on an X-ray film are used for assisting in determining the position of a focus of a patient or position and posture data of a surgical instrument in a surgical process, but some regions in a generated X-ray image have noise interference or are shielded by the marker balls (Mark balls), so that the observation of the X-ray image by a doctor is influenced, and therefore, defects such as the noise interference and the marker balls (Mark balls) shielding in the X-ray image need to be repaired by an image repairing method.
The aim of image restoration is to restore a lost area according to a certain rule by using known information in an image, and a Criminisi algorithm is a commonly used image restoration algorithm. The Criminisi algorithm selects a pixel point p with the highest priority on the edge of a region to be repaired, then constructs a pixel block with the size of n multiplied by n by taking p as the center, then searches a sample block which is most similar to the pixel block in the intact region, updates information to be repaired in the pixel block by using the found sample block, and starts the next iterative repair until the repair is completed. The Criminisi algorithm calculates the priority of the pixel point by using a formula P (P) = C (P) × D (P), wherein C (P) represents the confidence of a block to be repaired taking the pixel point P as the center, D (P) represents the intensity of an isolux line intersected with a boundary, C (P) may gradually become zero along with the repair of an image, the isolux line and a normal line in D (P) may have the condition of zero, and C (P) and D (P) may have two extremes of one extreme and one small extreme, which may cause the result of priority calculation to be inaccurate. In addition, the Criminisi algorithm is also fixed in size in the selection of sample blocks, size adjustment is not performed in combination with image information, and the Criminisi algorithm employs global search, which is time consuming.
For the defects that the priority calculation result deviation is caused by the extreme condition of the priority calculation of the pixel point of the existing Criminisi algorithm and the time complexity is higher in the image restoration process, a technical scheme for accurately and efficiently restoring the image is urgently needed.
Disclosure of Invention
The invention provides an image restoration method, an image restoration device, electronic equipment and a storage medium, which are used for solving the defects that in the prior art, the priority calculation result is not accurate enough due to extreme conditions of priority calculation of pixels of a Criminisi algorithm and the time complexity is high in the image restoration process, and realizing accurate and efficient restoration of an image.
The invention provides an image restoration method, which comprises the following steps:
determining the boundary of a current region to be repaired of a first image;
determining a current sample block size;
generating a first block to be repaired with a first pixel point as a center and a neighborhood block set adjacent to the first block to be repaired according to the size of the current sample block; the first pixel point is one of the pixel point sets on the boundary;
obtaining a correlation coefficient between the first block to be repaired and the neighborhood block set;
determining the priority of the first pixel point according to the correlation coefficient, the confidence of the first pixel point and the data item of the first pixel point;
determining a current pixel point to be repaired and a target matching block corresponding to the current pixel point to be repaired according to the priority of each pixel point in the pixel point set on the boundary, and repairing the block to be repaired corresponding to the current pixel point to be repaired according to the target matching block;
and updating the size of the sample block in the previous round according to the size of the initial sample block or the information entropy of the target matching block in the previous round of repair process.
According to an image restoration method provided by the present invention, the determining a boundary of a current area to be restored of a first image includes:
inputting the first image into a target detection model, and determining the boundary of the current to-be-repaired area of the first image; the target detection model is a pre-trained neural network model, a channel attention mechanism module is arranged on a feature fusion node of the target detection model, and the number of output nodes and the output size of the target detection model are preset according to the size feature of the region to be repaired in the first image.
According to an image restoration method provided by the present invention, the obtaining of the correlation coefficient between the first block to be restored and the neighborhood block set includes:
generating the correlation coefficient according to the variance of the first block to be repaired and the variance of each neighborhood block in the neighborhood block set; the neighborhood block set is composed of eight neighborhood blocks which are determined by grids with the same size as the first block to be repaired and are adjacent to the first block to be repaired at the central grid, and the eight neighborhood blocks do not contain pixel points of the area to be repaired.
According to the image restoration method provided by the invention, the correlation coefficient is obtained by the following formula:
wherein, R (p) represents the correlation coefficient of the first block to be repaired and the neighborhood block set, X represents the first block to be repaired, Y represents the neighborhood block in the neighborhood block set, var [ X ] represents the random quantity variance of X, var [ Y ] represents the random quantity variance of Y, cov (X, Y) represents the covariance of X and Y, and n is the number of the neighborhood blocks in the neighborhood block set.
According to the image inpainting method provided by the present invention, the determining the priority of the first pixel point according to the correlation coefficient, the confidence of the first pixel point and the data item of the first pixel point includes:
carrying out weighted summation on the correlation coefficient, the confidence coefficient of the first pixel point and the data item of the first pixel point to obtain the priority of the first pixel point; the correlation coefficient represents the degree of correlation between the image at the first pixel point and the image of the adjacent region, the confidence coefficient of the first pixel point represents the proportion of the image at the first pixel point containing known information, and the data item represents the linear complexity of the image structure at the first pixel point.
According to an image restoration method provided by the present invention, the determining a size of a current sample block includes:
under the condition that the current round is the first image restoration, acquiring the size of the initial sample block as the size of the current sample block; or, in the case where the current round is non-first image inpainting,
acquiring the information entropy of a target matching block in the previous round of repairing process;
increasing a current sample block size if the information entropy is less than a first threshold;
keeping the current sample block size unchanged if the information entropy is greater than or equal to the first threshold and less than or equal to a second threshold;
reducing the current sample block size if the information entropy is greater than the second threshold;
the value ranges of the first threshold and the second threshold are (0, 1), and the first threshold is smaller than the second threshold.
According to an image restoration method provided by the present invention, the determining, according to the priority of each pixel in the pixel set on the boundary, a current pixel to be restored and a target matching block corresponding to the current pixel to be restored, and restoring, according to the target matching block, a block to be restored corresponding to the current pixel to be restored, includes:
generating a plurality of image blocks within a first distance range of the current pixel point to be repaired, and determining a target matching block by calculating the similarity of the plurality of image blocks and a block to be repaired corresponding to the pixel point to be repaired;
repairing the block to be repaired corresponding to the current pixel point to be repaired according to the target matching block;
updating the first distance according to the similarity or the correlation coefficient of the target matching block and the block to be repaired corresponding to the pixel point to be repaired; wherein the initial value of the first distance is determined according to the distribution characteristics of the area to be repaired in the first image.
According to an image restoration method provided by the present invention, the updating the first distance according to the similarity or the correlation coefficient of the target matching block and the block to be restored corresponding to the pixel point to be restored includes:
and increasing the first distance when the similarity is smaller than a third threshold or the correlation coefficient is smaller than zero.
The present invention also provides an image restoration apparatus comprising:
the first determining module is used for determining the boundary of the current region to be repaired of the first image;
a second determining module, configured to determine a size of a current sample block;
a generating module, configured to generate, according to the size of the current sample block, a first block to be repaired with a first pixel point as a center, and a neighborhood block set adjacent to the first block to be repaired; wherein the first pixel point is one of the pixel point sets on the boundary;
an obtaining module, configured to obtain correlation coefficients of the first block to be repaired and the neighborhood block set;
a third determining module, configured to determine a priority of the first pixel according to the correlation coefficient, the confidence of the first pixel, and the data item of the first pixel;
the restoration module is used for determining the current pixel point to be restored and a target matching block corresponding to the current pixel point to be restored according to the priority of each pixel point in the pixel point set on the boundary, and restoring the block to be restored corresponding to the current pixel point to be restored according to the target matching block;
and updating the size of the sample block in the previous round according to the size of the initial sample block or the information entropy of the target matching block in the previous round of repair process.
The invention also provides an electronic device comprising a memory, a processor and a computer program stored on the memory and capable of running on the processor, wherein the processor executes the program to realize all or part of the steps of the image restoration method.
The present invention also provides a non-transitory computer readable storage medium having stored thereon a computer program which, when executed by a processor, implements all or part of the steps of any of the image inpainting methods described above.
According to the image restoration method, the image restoration device, the electronic equipment and the storage medium, the image restoration quality and efficiency are comprehensively improved through the size of the dynamic sample block, and the priority of the first pixel point is determined by introducing the correlation coefficient, combining the confidence of the first pixel point and the data item of the first pixel point, so that the obtained priority result is more comprehensive and accurate.
Drawings
In order to more clearly illustrate the technical solutions of the present invention or the prior art, the drawings needed for the description of the embodiments or the prior art will be briefly described below, and it is obvious that the drawings in the following description are some embodiments of the present invention, and those skilled in the art can also obtain other drawings according to the drawings without creative efforts.
FIG. 1 is a schematic flow chart of an image restoration method according to the present invention;
FIG. 2 is a schematic structural diagram of a channel attention mechanism module of a target detection model in an image restoration method according to the present invention;
FIG. 3 is a schematic structural diagram of a target detection model in an embodiment of an image restoration method provided by the present invention;
FIG. 4 is a schematic structural diagram of an image restoration apparatus according to the present invention;
fig. 5 is a schematic structural diagram of an electronic device provided in the present invention.
Detailed Description
In order to make the objects, technical solutions and advantages of the present invention clearer, the technical solutions of the present invention will be clearly and completely described below with reference to the accompanying drawings, and it is obvious that the described embodiments are some, but not all embodiments of the present invention. All other embodiments, which can be derived by a person skilled in the art from the embodiments given herein without making any creative effort, shall fall within the protection scope of the present invention.
An image restoration method, an image restoration apparatus, an electronic device, and a storage medium according to the present invention are described below with reference to fig. 1 to 5.
Fig. 1 is a schematic flow chart of an image restoration method according to the present invention, as shown in fig. 1, the method includes:
s11, determining the boundary of the current to-be-repaired area of the first image;
specifically, a better repairing effect can be obtained by gradually repairing the image from the edge of the region to be repaired, and for the first image to be repaired, the boundary of the region to be repaired needs to be determined. It can be understood that if there are multiple regions to be repaired, the regions need to be repaired separately, and the boundaries to be repaired are determined separately.
S12, determining the size of a current sample block; updating the size of the sample block in the previous round according to the size of the initial sample block or according to the information entropy of the target matching block in the previous round of repair;
specifically, in the prior art, the Criminisi algorithm generates a block to be repaired in a fixed sample block size, and then gradually starts repairing, in the invention, the size of the sample block is dynamic, and the size of the current sample block needs to be determined before the block to be repaired is generated in each round of repairing process.
And the size of the current sample block is determined by updating the size of the sample block in the previous round according to the size of the initial sample block or according to the information entropy of the target matching block in the previous round of repairing process. Namely, if the current time is the first time repair, the current sample block size is determined according to the preset initial sample block size, and if the current time is not the first time repair, the previous round of sample block size is updated according to the information entropy of the target matching block in the previous round of repair process and the preset rule to determine the current sample block size. The initial sample block size is set according to actual requirements or experience, for example, to 5 × 5, where "5 × 5" corresponds to a pixel block with 5 pixels in each row and each column. The target matching block is an image block determined from a good image area for repairing a block to be repaired, and the information entropy characterizes the amount of information in the target matching block.
S13, generating a first block to be repaired with a first pixel point as a center and a neighborhood block set adjacent to the first block to be repaired according to the size of the current sample block; wherein the first pixel point is one of the pixel point sets on the boundary;
specifically, taking the size of the current sample block 7 × 7 as an example, a pixel block with the size of 7 × 7 is determined on the first image with the first pixel point as the center to serve as a first block to be repaired, and then a plurality of pixel blocks adjacent to the first block to be repaired are generated on the first image with the same size to form a neighborhood block set of the first block to be repaired. The first pixel point is one of the pixel point sets on the boundary, that is, for the pixel points on the boundary of the current to-be-repaired area, the priority needs to be respectively determined, and the steps S13 to S15 are respectively executed. Of course, it can be understood that, for a pixel point that has not changed in the previous round of repair process, the priority of the pixel point in the previous round may be directly used without recalculation.
S14, obtaining correlation coefficients of the first block to be repaired and the neighborhood block set;
specifically, the neighborhood block set includes a plurality of neighborhood blocks, and the correlation between each neighborhood block and the first block to be repaired is calculated respectively, so as to obtain the correlation coefficient between the first block to be repaired and the neighborhood block set, where the correlation coefficient represents the similar correlation degree between the first block to be repaired and the surrounding area.
S15, determining the priority of the first pixel point according to the correlation coefficient, the confidence of the first pixel point and the data item of the first pixel point;
specifically, the correlation coefficient represents the degree of correlation between the image at the first pixel point and the image of the adjacent region, the confidence coefficient of the first pixel point represents the specific gravity of the image at the first pixel point containing known information, the data item represents the linear complexity of the image structure at the first pixel point, and the priority is obtained more comprehensively and accurately by integrating the correlation coefficient of the first to-be-repaired block of the first pixel point and the set of the adjacent blocks, the confidence coefficient of the first pixel point and the data item of the first pixel point.
S16, determining a current pixel point to be repaired and a target matching block corresponding to the current pixel point to be repaired according to the priority of each pixel point in the pixel point set on the boundary, and repairing the block to be repaired corresponding to the current pixel point to be repaired according to the target matching block;
specifically, after the priority of each pixel in the pixel set on the boundary is determined, the repairing sequence of the pixels can be determined, that is, the position on the boundary of the region to be repaired from which the pixel is to be repaired preferentially starts to be repaired. After the current pixel point to be repaired is determined, a matching block with the size of the sample block is further searched from the intact image area, a target matching block with the highest similarity of the block to be repaired corresponding to the current pixel point to be repaired is determined from the matching blocks, the block to be repaired corresponding to the current pixel point to be repaired is repaired according to the target matching block, and then the next round of repair is performed iteratively.
In the embodiment, the image restoration quality and efficiency are comprehensively improved through the size of the dynamic sample block, and the priority of the first pixel point is determined by introducing the correlation coefficient, combining the confidence coefficient of the first pixel point and the data item of the first pixel point, so that the obtained priority result is more comprehensive and accurate.
Based on any one of the above embodiments, in an embodiment, the determining the boundary of the current region to be repaired of the first image includes:
inputting the first image into a target detection model, and determining the boundary of the current to-be-repaired area of the first image; the target detection model is a pre-trained neural network model, a channel attention mechanism module is arranged on a feature fusion node of the target detection model, and the number of output nodes and the output size of the target detection model are preset according to the size feature of the region to be repaired in the first image.
Specifically, in this embodiment, a target detection model is used to detect a region to be repaired, so as to determine a boundary of the current region to be repaired, where the target detection model is a pre-trained neural network model. In the target detection model of this embodiment, a channel attention mechanism module is disposed at a feature fusion node, fig. 2 is a schematic structural diagram of the channel attention mechanism module, and as shown in fig. 2, in order to better calculate the importance degree of each feature, the channel attention mechanism module compresses an input feature graph F by Global Max Pooling (Global Max Pooling) and Global Average Pooling (Global Average Pooling) to obtain Global Max-Pooling feature graphs respectively
Global average pooled feature map
Adding element by element, sending the obtained result into a first full connection layer (FC) to construct the correlation among channels, simultaneously compressing the dimensionality, recovering the dimensionality into the initial dimensionality through a first activation function layer (ReLu) and a second full connection layer (FC), normalizing the calculated weight value to be between 0 and 1 through a second activation function layer (Sigmoid), and obtaining a channel attention weight characteristic diagram M c (F)∈R C×1×1 Finally, multiplying the feature calibration layer by the original input feature diagram F to obtain a new feature diagram F' = M c (F) And (F), outputting the image with the enhanced features by the feature fusion node according to the new feature graph, thereby accurately determining the area to be repaired.
In addition, in this embodiment, the number of output nodes and the output size of the target detection model are preset according to the size characteristic of the region to be repaired in the first image. For example, for an image to be repaired requiring the repair of a marker ball position, since the size of a marker ball (Mark ball) actually used is known in advance, for example, 5mm, the diameter of a marker ball area to be repaired in the image to be repaired is determined to be about 10 pixels, for this purpose, a neural network model is set to include two feature map output sizes 52 × 52 and 26 × 26, the prediction scale of the existing neural network model 13 × 13 is cancelled, fig. 3 is a schematic structural diagram of the corresponding neural network model, as shown in fig. 3, the output of the prediction scale of the existing neural network model 13 × 13 is removed, and only a feature-extracted 13 × 13 feature image is taken as reference information of an upper-level scale 26 × 26 image. Since the 13 × 13 prediction scale has a deep receptive field and is responsible for detecting a large target region, and the marker ball (Mark ball) in the input first image is small, the feature is not extracted in the deep network, so the 13 × 13 prediction scale output of the layer is removed and only used as the reference information of the upper layer scale.
In the embodiment, the image characteristics are strengthened through a channel attention mechanism, so that the result of detecting the current to-be-repaired area is more accurate, the number of output nodes and the output size of the target detection model are preset according to the size characteristics of the to-be-repaired area in the first image, and the efficiency of detecting the current to-be-repaired area is improved by using the priori knowledge of the known size characteristics of the to-be-repaired area.
Based on any one of the foregoing embodiments, in an embodiment, the obtaining the correlation coefficient between the first block to be repaired and the neighboring block set includes:
generating the correlation coefficient according to the variance of the first block to be repaired and the variance of each neighborhood block in the neighborhood block set; the neighborhood block set is composed of eight neighborhood blocks which are determined by grids with the same size as the first block to be repaired and are adjacent to the first block to be repaired at the central grid, and the eight neighborhood blocks do not contain pixel points of the area to be repaired.
Specifically, in a nine-square grid which takes a first block to be repaired as a central grid in a first image to be repaired, the size of eight peripheral grids is the same as that of the first block to be repaired at the central grid, neighborhood blocks which do not contain pixel points of a region to be repaired in the eight peripheral grids form a neighborhood block set, and a correlation coefficient between the first block to be repaired and the neighborhood block set is determined according to the variance of the first block to be repaired and the variance of each neighborhood block in the neighborhood block set.
In this embodiment, pixels in the region to be repaired are excluded in the determination process of the correlation number, and the size of the neighboring block is the same as that of the first block to be repaired, so that the variance of the first block to be repaired and the variance of each neighboring block in the set of neighboring blocks are comparable, and the correlation coefficient calculation is more accurate.
Based on any one of the above embodiments, in an embodiment, the correlation coefficient is obtained by the following formula:
wherein, R (p) represents the correlation coefficient of the first block to be repaired and the neighborhood block set, X represents the first block to be repaired, Y represents the neighborhood block in the neighborhood block set, var [ X ] represents the random quantity variance of X, var [ Y ] represents the random quantity variance of Y, cov (X, Y) represents the covariance of X and Y, and n is the number of the neighborhood blocks in the neighborhood block set.
In the embodiment, the correlation coefficient is accurately obtained by the formula.
Based on any one of the above embodiments, in an embodiment, the determining the priority of the first pixel according to the correlation coefficient, the confidence of the first pixel, and the data item of the first pixel includes:
carrying out weighted summation on the correlation coefficient, the confidence coefficient of the first pixel point and the data item of the first pixel point to obtain the priority of the first pixel point; the correlation coefficient represents the degree of correlation between the image at the first pixel point and the image of the adjacent region, the confidence coefficient of the first pixel point represents the proportion of the image at the first pixel point containing known information, and the data item represents the linear complexity of the image structure at the first pixel point.
Specifically, the priority of the first pixel point is obtained by performing weighted summation according to the correlation coefficient between the first block to be repaired of the first pixel point and the neighborhood block set, the confidence of the first pixel point, and the data item of the first pixel point, and accordingly, the weight may be manually preset or dynamically adjusted according to the actual image repairing effect. The calculation of the pixel point priority is more comprehensive by setting the weight, and the characteristics of specific types of images are matched, so that the image correction effect is improved
In this embodiment, the priority of the first pixel point is determined by performing weighted summation according to the correlation coefficient between the first block to be repaired of the first pixel point and the neighborhood block set, the confidence of the first pixel point, and the data item of the first pixel point, so that the priority calculation is more comprehensive and accurate, and the defect of misalignment of the priority of the pixel point under an extreme condition is avoided.
Based on any one of the above embodiments, in an embodiment, the priority of the first pixel point is obtained through the following formula:
P(p)＝a·C(p)+b·D(p)+c·R(p)；
wherein P is the first pixel point, P (P) is the priority of the first pixel point, a, b, c are weight values, a + b + c =1, and 0 ≦ a, b, c ≦ 1, c (P) is the confidence of the first pixel point, D (P) is the data item of the first pixel point, and R (P) is the correlation coefficient.
Based on any one of the above embodiments, in an embodiment, the confidence of the first pixel point is obtained through the following formula:
wherein p is the first pixel point, C (p) is the confidence of the first pixel point p, I represents the first image, Ω represents the region to be repaired, ψ p And representing the first block to be repaired. And q is one pixel point in the pixel point set which does not belong to the region to be repaired in the first block to be repaired.
Based on any one of the above embodiments, in an embodiment, the data item of the first pixel point is obtained through the following formula:
wherein p is the first pixel point, D (p) is a data item of the first pixel point,
indicating the direction of isoluminance, n p Represents the normal vector of pixel p, α =255 represents a normalization factor.
Based on the foregoing embodiments, in one embodiment, the determining the current sample block size includes:
under the condition that the current round is the first image restoration, acquiring the size of the initial sample block as the size of the current sample block; or, in the case where the current round is non-first image restoration,
acquiring the information entropy of a target matching block in the previous round of repairing process;
increasing a current sample block size if the information entropy is less than a first threshold;
keeping the current sample block size unchanged if the information entropy is greater than or equal to the first threshold and less than or equal to a second threshold;
reducing the current sample block size if the information entropy is greater than the second threshold;
the value ranges of the first threshold and the second threshold are (0, 1), and the first threshold is smaller than the second threshold.
Specifically, the initial sample block size is used as the current sample block size during the first round of image restoration, the current sample block size is increased when the information entropy is smaller than a first threshold value according to the information entropy of the target matching block during the previous round of image restoration, the current sample block size is kept unchanged when the information entropy is larger than or equal to the first threshold value and smaller than or equal to a second threshold value, and the current sample block size is decreased when the information entropy is larger than the second threshold value.
The information entropy of the image represents how much information is contained in the image, and is a statistical form of characteristics, which reflects how much information is contained in the image on average, and the one-dimensional information entropy of the image represents the information contained in the aggregation characteristics of the gray level distribution in the image. When the information entropy of the target sample block is smaller than a certain level, the image quality of the target sample block can be determined to be poor, and the size of the current sample block is increased so as to correspondingly increase the size of the target matching block of the current wheel and improve the image restoration quality; when the information entropy of the target sample block is larger than a certain level, the image quality redundancy of the target sample block can be determined, the size of the current sample block is reduced, the size of the target matching block of the current round is correspondingly reduced, and the image restoration efficiency is improved.
In the embodiment, the size of the current sample block is dynamically determined according to the information entropy of the target matching block in the previous round of repairing process, so that the pixel priority and the size of the target matching block are dynamically determined, and the image repairing quality and the image repairing efficiency are comprehensively improved.
Based on the above embodiments, in one embodiment, the current sample block size is obtained by the following formula:
wherein N represents the current nth iteration repair, and N n Represents the size of the sample block in the N-th round of the repair process, N n-1 Denotes the block size of the sample in the n-1 th round of the repair process, H n-1 Information entropy, K, representing target matching block in the n-1 th round of repair process 1 Representing said first threshold value, K 2 Representing the second threshold.
In this embodiment, the size of the current sample block is accurately obtained by the above formula.
Based on the above embodiments, in one embodiment, the information entropy of the target matching block is obtained according to the following formula:
wherein H is information entropy, i is gray value of pixel point, and P i Is the pixel dot out with the gray value iProbability now, assume P in the formula i Log (P) when =0 i )＝0。
Based on the foregoing embodiment, in an embodiment, the determining, according to the priority of each pixel in the pixel set on the boundary, a current pixel to be repaired and a target matching block corresponding to the current pixel to be repaired, and repairing, according to the target matching block, a block to be repaired corresponding to the current pixel to be repaired includes:
generating a plurality of image blocks within a first distance range of the current pixel point to be repaired, and determining a target matching block by calculating the similarity of the plurality of image blocks and a block to be repaired corresponding to the pixel point to be repaired;
repairing the block to be repaired corresponding to the current pixel point to be repaired according to the target matching block;
updating the first distance according to the similarity or the correlation coefficient of the target matching block and the block to be repaired corresponding to the pixel point to be repaired; wherein the initial value of the first distance is determined according to the distribution characteristics of the area to be repaired in the first image.
Specifically, the existing Criminisi algorithm adopts global search when determining pixel priority and a target matching block of a pixel point, and determines the target matching block in the whole image range, and the time consumption is large 0 For each pixel point, a first distance can be set to be L 0 . And determining an optimal target matching block according to the similarity of the plurality of image blocks generated by calculation and the to-be-repaired block corresponding to the to-be-repaired pixel point according to the first distance, and then repairing the to-be-repaired block corresponding to the current to-be-repaired pixel point according to the target matching block. After the repair, the first distance is updated according to the similarity of the target matching block and the block to be repaired corresponding to the pixel point to be repaired or the correlation coefficient of the current round, so as to achieve the aim of adjusting the repair quality according to the current roundThe matching range of the blocks to be repaired in the process of repairing is changed, and the image repairing efficiency and the image repairing quality are balanced. The similarity of the target matching block and the block to be repaired corresponding to the pixel point to be repaired can be mutual information similarity, histogram distance, barcol distance and the like.
In this embodiment, the initial matching range of the target matching block is determined according to the distribution characteristics of the to-be-repaired area in the first image, the initial image repairing efficiency is improved, and the effect of image repairing efficiency and image repairing quality is balanced by updating the first distance according to the similarity of the target matching block and the to-be-repaired block corresponding to the to-be-repaired pixel point or the correlation coefficient of the current round.
Based on the foregoing embodiment, in an embodiment, the updating the first distance according to the similarity or the correlation coefficient of the target matching block and the block to be repaired corresponding to the pixel point to be repaired includes: and increasing the first distance under the condition that the similarity of the target matching block and the block to be repaired corresponding to the pixel point to be repaired is smaller than a third threshold value or the correlation coefficient is smaller than zero.
Specifically, when the similarity of the target matching block and the block to be repaired corresponding to the pixel point to be repaired is smaller than a third threshold, it is indicated that the difference between the image of the target matching block and the block to be repaired is large, the image repairing quality is poor, and the target matching block with a higher fitting degree needs to be searched for to repair the block to be repaired, so that the first distance is increased, and the search matching range is expanded; the correlation coefficient is smaller than zero, which indicates that the difference between the image of the peripheral area of the first pixel point and the block to be repaired is larger, and a target matching block with higher fitting degree needs to be searched for to repair the block to be repaired, so that the first distance is increased, and the search matching range is expanded.
In this embodiment, the first distance is updated according to the similarity or the correlation coefficient of the target matching block and the block to be repaired corresponding to the pixel point to be repaired, so that the image repairing quality is improved.
Based on the above embodiments, in one embodiment, the first distance is updated according to the following formula:
where n denotes the current nth iterative repair, L n Representing a first distance in the n-th round of repair, R (p) representing the correlation coefficient, S representing the similarity of the target matching block to the block to be repaired corresponding to the pixel point to be repaired, and t representing the third threshold.
In the following, an image restoration apparatus provided by the present invention is described, and the image restoration apparatus described below and the image restoration method described above are referred to in correspondence with each other.
Fig. 4 is a schematic structural diagram of an image restoration apparatus according to the present invention, as shown in fig. 4, the apparatus includes:
a first determining module 41, configured to determine a boundary of a current region to be repaired of a first image;
a second determining module 42, configured to determine a current sample block size;
a generating module 43, configured to generate, according to the size of the current sample block, a first block to be repaired with a first pixel point as a center, and a neighborhood block set adjacent to the first block to be repaired; wherein the first pixel point is one of the pixel point sets on the boundary;
an obtaining module 44, configured to obtain correlation coefficients of the first block to be repaired and the neighborhood block set;
a third determining module 45, configured to determine a priority of the first pixel according to the correlation coefficient, the confidence of the first pixel, and the data item of the first pixel;
a repairing module 46, configured to determine a current pixel to be repaired and a target matching block corresponding to the current pixel to be repaired according to the priority of each pixel in the pixel set on the boundary, and repair a block to be repaired corresponding to the current pixel to be repaired according to the target matching block;
and updating the size of the sample block in the previous round according to the size of the initial sample block or the information entropy of the target matching block in the previous round of repair process.
In the embodiment, the image restoration quality and efficiency are comprehensively improved by the size of the dynamic sample block, and the priority of the first pixel point is determined by introducing the correlation coefficient and combining the confidence of the first pixel point and the data item of the first pixel point, so that the obtained priority result is more comprehensive and accurate.
Based on the foregoing embodiment, in an embodiment, the first determining module 41 is specifically configured to input the first image into the target detection model, and determine a boundary of a current region to be repaired of the first image; the target detection model is a pre-trained neural network model, a channel attention mechanism module is arranged on a feature fusion node of the target detection model, and the number of output nodes and the output size of the target detection model are preset according to the size feature of the region to be repaired in the first image.
In the embodiment, the image characteristics are strengthened through a channel attention mechanism, so that the result of detecting the current to-be-repaired area is more accurate, the number of output nodes and the output size of the target detection model are preset according to the size characteristics of the to-be-repaired area in the first image, and the efficiency of detecting the current to-be-repaired area is improved by using the priori knowledge of the known size characteristics of the to-be-repaired area.
Based on any of the foregoing embodiments, in an embodiment, the obtaining module 44 is configured to generate the correlation coefficient according to the variance of the first block to be repaired and the variance of each neighbor block in the neighbor block set; the neighborhood block set is composed of eight neighborhood blocks which are determined by grids with the same size as the first block to be repaired and are adjacent to the first block to be repaired at the central grid, and the eight neighborhood blocks do not contain pixel points of the area to be repaired.
In this embodiment, pixels in the region to be repaired are excluded in the determination process of the correlation number, and the size of the neighboring block is the same as that of the first block to be repaired, so that the variance of the first block to be repaired and the variance of each neighboring block in the set of neighboring blocks are comparable, and the correlation coefficient calculation is more accurate.
Based on any one of the above embodiments, in an embodiment, the third determining module 45 is configured to perform weighted summation on the correlation coefficient, the confidence level of the first pixel, and the data item of the first pixel, so as to obtain the priority of the first pixel; the correlation coefficient represents the degree of correlation between the image at the first pixel point and the image of an adjacent region, the confidence coefficient of the first pixel point represents the proportion of the image at the first pixel point containing known information, and the data item represents the linear complexity degree of the image structure at the first pixel point.
In this embodiment, the priority of the first pixel point is determined by performing weighted summation according to the correlation coefficient between the first block to be repaired of the first pixel point and the neighborhood block set, the confidence of the first pixel point, and the data item of the first pixel point, so that the priority calculation is more comprehensive and accurate, and the defect that the priority of the pixel point is inaccurate under extreme conditions is avoided.
Based on any of the above embodiments, in one embodiment, the second determining module 42 includes:
the first determining subunit is used for acquiring the size of the initial sample block as the size of the current sample block under the condition that the current round is the first image restoration;
the second determining subunit is used for acquiring the information entropy of the target matching block in the previous round of repairing process under the condition that the current round is the non-first-time image repairing; increasing a current sample block size if the information entropy is less than a first threshold; keeping the current sample block size unchanged if the information entropy is greater than or equal to the first threshold and less than or equal to a second threshold; in the event that the entropy of information is greater than the second threshold, decreasing the current block of samples size; the value ranges of the first threshold and the second threshold are (0, 1), and the first threshold is smaller than the second threshold.
In this embodiment, the size of the current sample block is dynamically determined according to the information entropy of the target matching block in the previous round of repairing process, so that the pixel priority and the size of the target matching block are dynamically determined, and the image repairing quality and the image repairing efficiency are comprehensively improved.
Based on any of the above embodiments, in one embodiment, the repair module 46 includes:
the first repairing subunit is used for generating a plurality of image blocks within a first distance range of the current pixel point to be repaired, and determining a target matching block by calculating the similarity of the plurality of image blocks and a block to be repaired corresponding to the pixel point to be repaired;
the second repairing subunit is used for repairing the to-be-repaired block corresponding to the current to-be-repaired pixel point according to the target matching block;
a third repairing subunit, configured to update the first distance according to the similarity or the correlation coefficient of the target matching block and the to-be-repaired block corresponding to the to-be-repaired pixel point; wherein the initial value of the first distance is determined according to the distribution characteristics of the area to be repaired in the first image.
In this embodiment, the initial matching range of the target matching block is determined according to the distribution characteristics of the to-be-repaired area in the first image, the initial image repairing efficiency is improved, and the effect of image repairing efficiency and image repairing quality is balanced by updating the first distance according to the similarity of the target matching block and the to-be-repaired block corresponding to the to-be-repaired pixel point or the correlation coefficient of the current round.
Based on any one of the foregoing embodiments, in an embodiment, the third repairing subunit is specifically configured to increase the first distance when the similarity is smaller than a third threshold or the correlation coefficient is smaller than zero.
In this embodiment, the first distance is updated according to the similarity or the correlation coefficient of the target matching block and the block to be repaired corresponding to the pixel point to be repaired, so that the image repairing quality is improved.
Fig. 5 illustrates a physical structure diagram of an electronic device, which may include, as shown in fig. 5: a processor (processor) 510, a communication Interface (Communications Interface) 520, a memory (memory) 530, and a communication bus 540, wherein the processor 510, the communication Interface 520, and the memory 530 communicate with each other via the communication bus 540. Processor 510 may invoke logic instructions in memory 530 to perform all or a portion of the steps of the image inpainting methods provided above, including: determining the boundary of a current region to be repaired of a first image; determining a current sample block size; generating a first block to be repaired with a first pixel point as a center and a neighborhood block set adjacent to the first block to be repaired according to the size of the current sample block; the first pixel point is one of the pixel point sets on the boundary; obtaining a correlation coefficient between the first block to be repaired and the neighborhood block set; determining the priority of the first pixel point according to the correlation coefficient, the confidence of the first pixel point and the data item of the first pixel point; determining a current pixel point to be repaired and a target matching block corresponding to the current pixel point to be repaired according to the priority of each pixel point in the pixel point set on the boundary, and repairing the block to be repaired corresponding to the current pixel point to be repaired according to the target matching block; and updating the size of the sample block in the previous round according to the size of the initial sample block or the information entropy of the target matching block in the previous round of repair process.
In addition, the logic instructions in the memory 830 may be implemented in software functional units and stored in a computer readable storage medium when the logic instructions are sold or used as independent products. Based on such understanding, the technical solution of the present invention may be embodied in the form of a software product, which is stored in a storage medium and includes instructions for causing a computer device (which may be a personal computer, a server, or a network device) to execute all or part of the steps of the method according to the embodiments of the present invention. And the aforementioned storage medium includes: a U-disk, a removable hard disk, a Read-Only Memory (ROM), a Random Access Memory (RAM), a magnetic disk or an optical disk, and other various media capable of storing program codes.
In another aspect, the present invention also provides a computer program product comprising a computer program stored on a non-transitory computer readable storage medium, the computer program comprising program instructions which, when executed by a computer, enable the computer to perform all or part of the steps of the image inpainting method provided above, the method comprising: determining the boundary of a current region to be repaired of a first image; determining a current sample block size; generating a first block to be repaired with a first pixel point as a center and a neighborhood block set adjacent to the first block to be repaired according to the size of the current sample block; the first pixel point is one of the pixel point sets on the boundary; obtaining a correlation coefficient between the first block to be repaired and the neighborhood block set; determining the priority of the first pixel point according to the correlation coefficient, the confidence of the first pixel point and the data item of the first pixel point; determining a current pixel point to be repaired and a target matching block corresponding to the current pixel point to be repaired according to the priority of each pixel point in the pixel point set on the boundary, and repairing the block to be repaired corresponding to the current pixel point to be repaired according to the target matching block; and updating the size of the sample block in the previous round according to the size of the initial sample block or according to the information entropy of the target matching block in the previous round of repair.
In yet another aspect, the present invention also provides a non-transitory computer-readable storage medium having stored thereon a computer program, which when executed by a processor is implemented to perform all or part of the steps of the image inpainting method provided above, the method including: determining the boundary of a current region to be repaired of a first image; determining a current sample block size; generating a first block to be repaired with a first pixel point as a center and a neighborhood block set adjacent to the first block to be repaired according to the size of the current sample block; the first pixel point is one of the pixel point sets on the boundary; obtaining a correlation coefficient between the first block to be repaired and the neighborhood block set; determining the priority of the first pixel point according to the correlation coefficient, the confidence of the first pixel point and the data item of the first pixel point; determining a current pixel point to be repaired and a target matching block corresponding to the current pixel point to be repaired according to the priority of each pixel point in the pixel point set on the boundary, and repairing the block to be repaired corresponding to the current pixel point to be repaired according to the target matching block; and updating the size of the sample block in the previous round according to the size of the initial sample block or the information entropy of the target matching block in the previous round of repair process.
The above-described embodiments of the apparatus are merely illustrative, and the units described as separate parts may or may not be physically separate, and parts displayed as units may or may not be physical units, may be located in one place, or may be distributed on a plurality of network units. Some or all of the modules may be selected according to actual needs to achieve the purpose of the solution of the present embodiment. One of ordinary skill in the art can understand and implement it without inventive effort.
Through the above description of the embodiments, those skilled in the art will clearly understand that each embodiment may be implemented by software plus a necessary general hardware platform, and may also be implemented by hardware. With this understanding in mind, the above-described technical solutions may be embodied in the form of a software product, which can be stored in a computer-readable storage medium such as ROM/RAM, magnetic disk, optical disk, etc., and includes instructions for causing a computer device (which may be a personal computer, a server, or a network device, etc.) to execute the methods described in the embodiments or some parts of the embodiments.
Finally, it should be noted that: the above examples are only intended to illustrate the technical solution of the present invention, and not to limit it; although the present invention has been described in detail with reference to the foregoing embodiments, it will be understood by those of ordinary skill in the art that: the technical solutions described in the foregoing embodiments may still be modified, or some technical features may be equivalently replaced; and such modifications or substitutions do not depart from the spirit and scope of the corresponding technical solutions of the embodiments of the present invention.
Claims (11)
1. An image restoration method, comprising:
determining the boundary of a current region to be repaired of a first image;
determining a current sample block size;
generating a first block to be repaired taking a first pixel point as a center and a neighborhood block set adjacent to the first block to be repaired according to the size of the current sample block; wherein the first pixel point is one of the pixel point sets on the boundary;
obtaining a correlation coefficient between the first block to be repaired and the neighborhood block set;
determining the priority of the first pixel point according to the correlation coefficient, the confidence of the first pixel point and the data item of the first pixel point;
determining a current pixel point to be repaired and a target matching block corresponding to the current pixel point to be repaired according to the priority of each pixel point in the pixel point set on the boundary, and repairing the block to be repaired corresponding to the current pixel point to be repaired according to the target matching block;
and updating the size of the sample block in the previous round according to the size of the initial sample block or according to the information entropy of the target matching block in the previous round of repair.
2. The image inpainting method of claim 1, wherein the determining the boundary of the current area to be inpainted of the first image comprises:
inputting the first image into a target detection model, and determining the boundary of the current to-be-repaired area of the first image; the target detection model is a pre-trained neural network model, a channel attention mechanism module is arranged on a feature fusion node of the target detection model, and the number of output nodes and the output size of the target detection model are preset according to the size feature of the region to be repaired in the first image.
3. The image inpainting method according to claim 1, wherein the obtaining of the correlation coefficient between the first block to be repaired and the neighborhood block set comprises:
generating the correlation coefficient according to the variance of the first block to be repaired and the variance of each neighborhood block in the neighborhood block set; the neighborhood block set is composed of eight neighborhood blocks which are determined by grids with the same size as the first block to be repaired and are adjacent to the first block to be repaired at the central grid, and the eight neighborhood blocks do not contain pixel points of the area to be repaired.
4. The image inpainting method of claim 1, wherein the correlation coefficient is obtained by the following formula:
wherein, R (p) represents the correlation coefficient of the first block to be repaired and the neighborhood block set, X represents the first block to be repaired, Y represents the neighborhood block in the neighborhood block set, var [ X ] represents the random quantity variance of X, var [ Y ] represents the random quantity variance of Y, cov (X, Y) represents the covariance of X and Y, and n is the number of the neighborhood blocks in the neighborhood block set.
5. The image inpainting method of claim 1, wherein the determining the priority of the first pixel according to the correlation coefficient, the confidence of the first pixel, and the data item of the first pixel comprises:
carrying out weighted summation on the correlation coefficient, the confidence coefficient of the first pixel point and the data item of the first pixel point to obtain the priority of the first pixel point; the correlation coefficient represents the degree of correlation between the image at the first pixel point and the image of the adjacent region, the confidence coefficient of the first pixel point represents the proportion of the image at the first pixel point containing known information, and the data item represents the linear complexity of the image structure at the first pixel point.
6. The image inpainting method of claim 1, wherein the determining a current sample block size comprises:
under the condition that the current round is the first image restoration, acquiring the size of the initial sample block as the size of the current sample block; or, in the case where the current round is non-first image restoration,
acquiring the information entropy of a target matching block in the previous round of repairing process;
increasing a current sample block size if the information entropy is less than a first threshold;
keeping the current sample block size unchanged if the information entropy is greater than or equal to the first threshold and less than or equal to a second threshold;
reducing the current sample block size if the information entropy is greater than the second threshold;
the value ranges of the first threshold and the second threshold are (0, 1), and the first threshold is smaller than the second threshold.
7. The image inpainting method according to claim 1, wherein the determining, according to the priority of each pixel in the pixel set on the boundary, a current pixel to be inpainted and a target matching block corresponding to the current pixel to be inpainted, and inpainting, according to the target matching block, an inpainting block corresponding to the current pixel to be inpainted comprises:
generating a plurality of image blocks within a first distance range of the current pixel point to be repaired, and determining a target matching block by calculating the similarity of the plurality of image blocks and the block to be repaired corresponding to the pixel point to be repaired;
repairing the block to be repaired corresponding to the current pixel point to be repaired according to the target matching block;
updating the first distance according to the similarity or the correlation coefficient of the target matching block and the block to be repaired corresponding to the pixel point to be repaired; wherein the initial value of the first distance is determined according to the distribution characteristics of the area to be repaired in the first image.
8. The image inpainting method according to claim 7, wherein the updating the first distance according to the similarity or the correlation coefficient of the block to be inpainted corresponding to the pixel point to be inpainted by the target matching block comprises:
and increasing the first distance when the similarity is smaller than a third threshold or the correlation coefficient is smaller than zero.
9. An image restoration apparatus, comprising:
the first determining module is used for determining the boundary of the current region to be repaired of the first image;
a second determining module, configured to determine a size of a current sample block;
a generating module, configured to generate, according to the size of the current sample block, a first block to be repaired with a first pixel point as a center, and a neighborhood block set adjacent to the first block to be repaired; the first pixel point is one of the pixel point sets on the boundary;
an obtaining module, configured to obtain correlation coefficients of the first block to be repaired and the neighborhood block set;
a third determining module, configured to determine a priority of the first pixel according to the correlation coefficient, the confidence of the first pixel, and the data item of the first pixel;
the restoration module is used for determining a current pixel point to be restored and a target matching block corresponding to the current pixel point to be restored according to the priority of each pixel point in the pixel point set on the boundary, and restoring the block to be restored corresponding to the current pixel point to be restored according to the target matching block;
and updating the size of the sample block in the previous round according to the size of the initial sample block or the information entropy of the target matching block in the previous round of repair process.
10. An electronic device comprising a memory, a processor and a computer program stored on the memory and executable on the processor, wherein the processor implements all or part of the steps of the document detection method according to any one of claims 1 to 8 when executing the program.
11. A non-transitory computer readable storage medium having stored thereon a computer program, wherein the computer program, when executed by a processor, implements all or part of the steps of the document detection method according to any one of claims 1 to 8.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202111015407.8A CN115797186A (en) | 2021-08-31 | 2021-08-31 | Image restoration method and device, electronic equipment and storage medium |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202111015407.8A CN115797186A (en) | 2021-08-31 | 2021-08-31 | Image restoration method and device, electronic equipment and storage medium |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN115797186A true CN115797186A (en) | 2023-03-14 |
Family
ID=85473413
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202111015407.8A Pending CN115797186A (en) | 2021-08-31 | 2021-08-31 | Image restoration method and device, electronic equipment and storage medium |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN115797186A (en) |
Cited By (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN117197730A (en) * | 2023-11-02 | 2023-12-08 | 江苏通创现代建筑产业技术研究院有限公司 | Repair evaluation method for urban space distortion image |
-
2021
- 2021-08-31 CN CN202111015407.8A patent/CN115797186A/en active Pending
Cited By (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN117197730A (en) * | 2023-11-02 | 2023-12-08 | 江苏通创现代建筑产业技术研究院有限公司 | Repair evaluation method for urban space distortion image |
| CN117197730B (en) * | 2023-11-02 | 2024-04-05 | 江苏通创现代建筑产业技术研究院有限公司 | Repair evaluation method for urban space distortion image |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN110321920B (en) | Image classification method and device, computer readable storage medium and computer equipment | |
| US9697602B1 (en) | System and method for auto-contouring in adaptive radiotherapy | |
| CN110276745B (en) | Pathological image detection algorithm based on generation countermeasure network | |
| US8873826B2 (en) | Method for brightness level calculation of the digital x-ray image for medical applications | |
| CN109480780B (en) | Evaluation method and system of stroke early warning system | |
| CN108364297B (en) | Blood vessel image segmentation method, terminal and storage medium | |
| WO2021136368A1 (en) | Method and apparatus for automatically detecting pectoralis major region in molybdenum target image | |
| WO2020234349A1 (en) | Sampling latent variables to generate multiple segmentations of an image | |
| CN111695624A (en) | Data enhancement strategy updating method, device, equipment and storage medium | |
| CN111462102A (en) | Intelligent analysis system and method based on novel coronavirus pneumonia X-ray chest radiography | |
| CN111080592B (en) | Rib extraction method and device based on deep learning | |
| CN111626379A (en) | X-ray image detection method for pneumonia | |
| CN115797186A (en) | Image restoration method and device, electronic equipment and storage medium | |
| CN111862071B (en) | Method for measuring CT value of lumbar 1 vertebral body based on CT image | |
| EP4168931A1 (en) | Domain aware medical image classifier interpretation by counterfactual impact analysis | |
| CN113256670A (en) | Image processing method and device, and network model training method and device | |
| CN112967293A (en) | Image semantic segmentation method and device and storage medium | |
| CN111126424B (en) | Ultrasonic image classification method based on convolutional neural network | |
| CN108447066B (en) | Biliary tract image segmentation method, terminal and storage medium | |
| CN116246126A (en) | Iterative unsupervised domain self-adaption method and device | |
| CN115439423B (en) | CT image-based identification method, device, equipment and storage medium | |
| CN116051421A (en) | Multi-dimensional-based endoscope image quality evaluation method, device, equipment and medium | |
| US20220414869A1 (en) | Detecting and segmenting regions of interest in biomedical images using neural networks | |
| CN116309806A (en) | CSAI-Grid RCNN-based thyroid ultrasound image region of interest positioning method | |
| CN112634224B (en) | Focus detection method and device based on target image |
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 |