Image preprocessing system for image processing

Abstract

The invention discloses an image pre-processing system for image processing, which relates to the technical field of image processing, and comprises the following steps of: performing initial state processing on a target image, performing noise reduction on the target image by virtue of a filter, performing edge removal on the target image, obtaining a plurality of images with marked edge parts, marking the images as comparison images, and performing image processing on the comparison images. The marked edge part refers to a part without an actual image; a plurality of cutting distances are determined in the comparison image, and one cutting distance is determined as an acquisition distance according to the distribution condition of the cutting distances, so that the target image is removed, and a key part is left; and then the removed data are identified by means of an intelligent identification model, and a result is output. The method is simple, effective, easy and practical.

Description

Technical Field

[0001] The invention belongs to the technical field of image processing, and in particular to an image pre-processing system for image processing. Background Art

[0002] The patent with publication number CN117649410A discloses a medical image recognition technology, specifically a medical image processing method. This application extracts the frequency domain features and spatial domain features of the image to be processed, and fuses the above two features into fusion features. The first recognition result and the second recognition result are obtained by taking the fusion features and the spatial domain features as recognition feature objects, and then the first recognition result and the second recognition result are fused and integrated according to the preset weights to obtain the final recognition result. This method improves the accuracy of medical image detection through the correlation between the frequency domain and the spatial domain.

[0003] For the target image, how to perform reasonable pre-processing, divide the key parts, and then use the key parts to perform basic processing on the image; based on this, a solution is now provided. Summary of the invention

[0004] The present invention aims to solve at least one of the technical problems existing in the prior art;

[0005] To this end, the present invention proposes an image pre-processing system for image processing, comprising:

[0006] The initial processing unit is used to perform initial processing on the target image. The specific method of initial processing is as follows:

[0007] The target image is acquired, and then the target image is subjected to noise reduction by means of a filter;

[0008] Then the target image is edge removed. The specific method of edge removal is as follows:

[0009] A number of images with marked edge parts are obtained and marked as comparison images, where the marked edge parts refer to parts without actual images;

[0010] Acquire several acquisition points on the outermost edge line of the comparison image. The number of acquisition points is set by the administrator.

[0011] Then, the shortest distance between the acquisition point on each edge line and another edge line inside the edge part is obtained, and the distance is marked as the cutting distance;

[0012] Select any acquisition point and obtain several cutting distances Gi, i = 1, ..., n of the acquisition point, which is represented by the presence of n labeled contrast images corresponding to the edge parts. Then, the mean value P of Gi is automatically obtained, and the steady-state value W of Gi is calculated using the formula. The specific calculation formula is:

[0013]

[0014] When W exceeds the set value X2, Gi is sorted in descending order of |Gi-P|, and one Gi is selected in turn. Each selected one is deleted, and the steady-state value W of the remaining Gi is calculated after deletion. If W still exceeds X2, the next Gi is automatically selected until the steady-state value W of the remaining Gi does not exceed X2, and the mean value of the remaining Gi is marked as the collection distance;

[0015] Then, the remaining collection points are processed in the same way to obtain the deemed collection distances corresponding to all collection points;

[0016] Then, several real-time acquisition points are selected on the target image according to the same point positions, and each real-time acquisition point is narrowed down to the inside of the target image to determine several internal points corresponding to the acquisition points;

[0017] Connect adjacent internal points in sequence, mark the connected points as inner edge lines, and mark the parts outside the inner edge lines of the target image as cut parts;

[0018] The cut portion is removed from the target image, and the remaining is marked as the approved image.

[0019] Furthermore, a median filtering method is used to filter the target image.

[0020] Furthermore, the collection points are determined in such a manner that the difference between the arc length and the straight line length between the two collection points does not exceed a preset value X1.

[0021] Furthermore, a plurality of acquisition points are obtained on the outermost edge line of the comparison image in a manner of evenly dividing the edge arc length.

[0022] Furthermore, the narrowing method is to select the corresponding internal point in a manner of extending the longest point inside the target image according to the collection distance corresponding to each real-time collection point.

[0023] Furthermore, it also includes an image acquisition unit, which is used to acquire all target images, and the target images are referred to as medical images in this application;

[0024] The image acquisition unit is used to transmit the target image to the primary processing unit.

[0025] Furthermore, the data source collection unit is used to collect a number of abnormal part images corresponding to various target images that have been annotated, including abnormal part images, abnormal points, and abnormal reasons corresponding to the abnormal points, and divide the abnormal part images into a training set and a verification set, with a division ratio of seven to three;

[0026] The training set is then put into the intelligent recognition model for training. After the training, the recognition result is verified by the verification set. If the accuracy of the recognition result exceeds the set ratio B1, B1 is a preset value, it means that the training of the intelligent recognition model is completed and the intelligent recognition model is obtained.

[0027] Furthermore, the model recognition unit is used to transmit the intelligent recognition model to the processor, and the initial processing unit is used to transmit the approved image to the processor.

[0028] Furthermore, the processor is used to identify the approved image with the help of an intelligent recognition model, identify abnormal points, and obtain abnormal results.

[0029] Furthermore, the processor is used to transmit the abnormal result to the output unit, and at the same time, the abnormal result is timestamped and transmitted to the storage unit for real-time storage.

[0030] Compared with the prior art, the present invention has the following beneficial effects:

[0031] The present application performs initial state processing on the target image, removes the edge of the target image after denoising the target image with the help of a filter, obtains a number of images with marked edge portions, marks them as comparison images, and the marked edge portions refer to portions without actual images;

[0032] When comparing images, several cutting distances are determined, and a distance deemed as a collection distance is determined according to the distribution of the cutting distances, so as to remove the target image and leave the key part; then the removed part is identified with the help of an intelligent recognition model and the result is output; the present invention is simple, effective and easy to use. BRIEF DESCRIPTION OF THE DRAWINGS

[0033] Figure 1 It is a system block diagram of the present invention. DETAILED DESCRIPTION

[0034] The technical solution of the present invention will be clearly and completely described below in conjunction with the embodiments. Obviously, the described embodiments are only part of the embodiments of the present invention, rather than all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0035] See also Figure 1 , the present application provides an image pre-processing system for image processing, comprising:

[0036] An image acquisition unit, used to acquire all target images, where the target images in this application refer to medical images, specifically including one or more of X-ray images, CT images, ultrasound images, DSA angiography images, and endoscopic examination images;

[0037] The image acquisition unit is used to transmit the target image to the initial processing unit, and the initial processing unit is used to perform initial state processing on the target image. The specific method of initial state processing is:

[0038] The target image is acquired, and then the target image is subjected to noise reduction by means of a filter. Specifically, the target image is filtered by a median filter to remove the noise in the target image.

[0039] Then the target image is edge removed. The specific method of edge removal is as follows:

[0040] Obtain several images with marked edge parts, mark them as comparison images, and the marked edge parts refer to the parts without actual images. The comparison images here are all processed by the same operator;

[0041] Then, a number of acquisition points are obtained on the outermost edge line of the comparison image by evenly dividing the edge arc length. The number of acquisition points is set by the administrator. Here, the acquisition can also be obtained by a method in which the difference between the arc length and the straight line length between the two acquisition points does not exceed a preset value X1;

[0042] Then, the shortest distance between the acquisition point on each edge line and another edge line inside the edge part is obtained, and the distance is marked as the cutting distance;

[0043] Select any acquisition point and obtain several cutting distances Gi, i = 1, ..., n of the acquisition point, which is represented by the presence of n labeled contrast images corresponding to the edge parts. Then, the mean value P of Gi is automatically obtained, and the steady-state value W of Gi is calculated using the formula. The specific calculation formula is:

[0044]

[0045] When W exceeds the set value X2, Gi is sorted in descending order of |Gi-P|, and one Gi is selected in turn. Each selected one is deleted, and the steady-state value W of the remaining Gi is calculated after deletion. If W still exceeds X2, the next Gi is automatically selected until the steady-state value W of the remaining Gi does not exceed X2, and the mean value of the remaining Gi is marked as the collection distance;

[0046] Then, the remaining collection points are processed in the same way to obtain the deemed collection distances corresponding to all collection points;

[0047] Then, several real-time acquisition points are selected on the target image according to the same point position, and each real-time acquisition point is narrowed down to the inside of the target image. The narrowing method is to select the corresponding internal point in the way of extending the longest point inside the target image according to the acquisition distance corresponding to each real-time acquisition point;

[0048] Connect adjacent internal points in sequence, mark the connected points as inner edge lines, and mark the parts outside the inner edge lines of the target image as cut parts;

[0049] The cut part is removed from the target image, and the remaining part is marked as the approved image;

[0050] The primary processing unit is used to transmit the approved image to the processor;

[0051] The data source collection unit is used to collect a number of abnormal part images corresponding to various target images that have been annotated, including abnormal part images, abnormal points, and abnormal reasons corresponding to the abnormal points, and divide the abnormal part images into training sets and verification sets, with a division ratio of seven to three, or set by the administrator; the data source collection unit is used to transmit the training set and the verification set to the model recognition unit, and the model recognition unit is used to train the recognition model. The specific training method is:

[0052] Then the training set is put into the intelligent recognition model for training. After the training, the recognition result is verified by the verification set. If the accuracy of the recognition result exceeds the set ratio B1, B1 is a preset value, it means that the training of the intelligent recognition model is completed. The intelligent recognition model here is the convolutional neural network model CNN. Of course, other neural network models can also be used.

[0053] Obtain an intelligent recognition model;

[0054] The model recognition unit is used to transmit the intelligent recognition model to the processor, and the processor is used to recognize the approval image with the help of the intelligent recognition model, identify the abnormal points, and obtain the abnormal result;

[0055] The processor is used to transmit the abnormal result to the output unit, and at the same time, the abnormal result is timestamped and transmitted to the storage unit for real-time storage.

[0056] The above embodiments are only used to illustrate the technical method of the present invention rather than to limit it. Although the present invention has been described in detail with reference to the preferred embodiments, those skilled in the art should understand that the technical method of the present invention may be modified or replaced by equivalents without departing from the spirit and scope of the technical method of the present invention.

Claims

1. An image pre-processing system for image processing, characterized in that: include: The initial processing unit is used to perform initial processing on the target image. The specific method of initial processing is as follows: The target image is acquired, and then the target image is subjected to noise reduction by means of a filter; Then the target image is edge removed. The specific method of edge removal is as follows: A number of images with marked edge parts are obtained and marked as comparison images, where the marked edge parts refer to parts without actual images; Acquire several acquisition points on the outermost edge line of the comparison image. The number of acquisition points is set by the administrator. Then, the shortest distance between the acquisition point on each edge line and another edge line inside the edge part is obtained, and the distance is marked as the cutting distance; Select any acquisition point and obtain several cutting distances Gi of the acquisition point, i = 1, ..., n, which means there are n labeled contrast images corresponding to the edge parts. Then automatically obtain the mean value P of Gi, and use the formula to calculate the steady-state value W of Gi. The specific calculation formula is: When W exceeds the set value X2, Gi is sorted in descending order of |Gi-P|, and one Gi is selected in turn. Each selected one is deleted, and the steady-state value W of the remaining Gi is calculated after deletion. If W still exceeds X2, the next Gi is automatically selected until the steady-state value W of the remaining Gi does not exceed X2, and the mean value of the remaining Gi is marked as the collection distance; Then, the remaining collection points are processed in the same way to obtain the deemed collection distances corresponding to all collection points; Then, several real-time acquisition points are selected on the target image according to the same point positions, and each real-time acquisition point is narrowed down to the inside of the target image to determine several internal points corresponding to the acquisition points; Connect adjacent internal points in sequence, mark the connected points as inner edge lines, and mark the parts outside the inner edge lines of the target image as cut parts; The cut portion is removed from the target image, and the remaining is marked as the approved image.

2. The image pre-processing system for image processing according to claim 1, characterized in that: The median filtering method is used to filter the target image.

3. The image pre-processing system for image processing according to claim 1, characterized in that: The collection points are determined in the following manner: the difference between the arc length and the straight line length between the two collection points does not exceed a preset value X1.

4. The image pre-processing system for image processing according to claim 1, characterized in that: Several acquisition points are obtained on the outermost edge line of the comparison image by evenly dividing the edge arc length.

5. The image pre-processing system for image processing according to claim 1, characterized in that: The limiting method is to select the corresponding internal point by extending the longest point inside the target image according to the collection distance corresponding to each real-time collection point.

6. The image pre-processing system for image processing according to claim 1, characterized in that: It also includes an image acquisition unit for acquiring all target images, where the target images are referred to as medical images in this application; The image acquisition unit is used to transmit the target image to the primary processing unit.

7. The image pre-processing system for image processing according to claim 1, characterized in that: The data source collection unit is used to collect a number of abnormal part images corresponding to various target images that have been annotated, including abnormal part images, abnormal points, and abnormal reasons corresponding to the abnormal points, and divide the abnormal part images into a training set and a verification set, with a division ratio of seven to three; The training set is then put into the intelligent recognition model for training. After the training, the recognition result is verified by the verification set. If the accuracy of the recognition result exceeds the set ratio B1, B1 is a preset value, it means that the training of the intelligent recognition model is completed and the intelligent recognition model is obtained.

8. The image pre-processing system for image processing according to claim 1, characterized in that: The model recognition unit is used to transmit the intelligent recognition model to the processor, and the initial processing unit is used to transmit the approved image to the processor.

9. The image pre-processing system for image processing according to claim 1, characterized in that: The processor is used to identify the approved image with the help of an intelligent recognition model, identify abnormal points, and obtain abnormal results.

10. The image pre-processing system for image processing according to claim 1, characterized in that: The processor is used to transmit the abnormal result to the output unit, and at the same time, the abnormal result is timestamped and transmitted to the storage unit for real-time storage.

Images