Multiscale level set image segmenting method based on kernel fuzzy clustering

Abstract

The invention discloses a multiscale level set image segmenting method based on kernel fuzzy clustering. The method includes step 1 adopting a divide marking method to calculate the gray average of small areas; 2 adopting the gray average to initialize a membership matrix, conducting kernel fuzzy clustering to obtain the initial contour of the interested area; 3 designing an edge constraint stop item of the multiscale level set; 4 conducting iteration evolution to segment an image. The method overcomes the shortcoming that over-segmenting is easy to cause for the divide based on edge information, and edge segmenting missing is easy to cause for the C-V model level set method based on area information for the images with unclear edges and poor contrast. The edge information and the area information are effectively blended by adopting the kernel fuzzy clustering method, the multiscale edge constraint stop item is added, the re-initialization is eliminated, the segmenting accuracy is improved, and the real-time performance of the algorithm is ensured.

Description

technical field [0001] The invention belongs to the field of image processing, in particular based on a multi-scale level set image segmentation method based on kernel fuzzy clustering. Background technique [0002] Image processing technology is widely used in military, medical, industrial production and other fields, and the segmentation effect of the target plays a vital role in the completion of related tasks. Therefore, the method with good real-time performance and high segmentation accuracy has always been a research hotspot of many scholars. The traditional image segmentation method is either based on edge information or based on region information. It is not very ideal for image segmentation with blurred edges, poor contrast, and uneven gray scale. Although scholars have also proposed many methods that combine edge and region information. method, but it is still difficult to balance real-time and accuracy. [0003] In recent years, the level set method has become ...

AI Technical Summary

Problems solved by technology

The traditional image segmentation method is either based on edge information or based on region information. It is not very ideal for image segmentation with blurred edges, poor contrast, and uneven gray scale. Although scholars have also proposed many methods that combine edge and region information. method, but it is still difficult to balance real-time and accuracy

[0003] In recent years, the level set method has become one of the main development directions in the field of image segmentation, but it also has some limitations: the evolution process needs to be re-initialized, and images with uneven gray levels, poor edge contrast, and complex edges are prone to missing segmentation or over-segmentation. Segmentation phenomenon

The existing level set image segmentation methods mostly use manual or semi-automatically to provide the initial level set for segmentation, which reduces the degree of intelligence; Li Chunming et al. study the symbolic distance function to eliminate re-initialization, but use edge information for image segmentation, which is difficult for images with uneven gray levels. The segmentation effect is not ideal; although some level set methods that integrate edge and area information have played a certain role in the segmentation effect, they do not make full use of the image gray level information and ingenious design to obtain the initial contour, which affects the level set segmentation. Timeliness, Intelligence and Accuracy

Images