Image Segmentation Adaptive Distance Preserving Level Set Evolution Analysis Based on Depth Learning
Download as PDF
DOI: 10.25236/iwmecs.2020.043
Author(s)
Liping Wang, Yinhua Wang
Corresponding Author
Liping Wang
Abstract
Image segmentation is an important and low-level processing task in computer vision. The quality of segmentation has a great impact on the completion of subsequent tasks. From the perspective of visual perception, image segmentation seems to be a very easy problem to solve. In actual image processing, it is a very difficult task. Compared with image classification and target detection, distance-preserving level set method involves two cross-cutting fields of image recognition and natural language processing, thus it is extremely challenging. In the research and application of computer vision, image segmentation is often the first step and plays a very important role in the whole process. The zero level set curve can adaptively decide whether to move inward or outward according to the characteristics of the image, and the curve can continue to evolve to the boundary of the target object in the region where the gray values of the pixels are equal. Adaptive distance keeping level set evolution model introduces variable weight coefficient on the basis of no need to initialize the model, so as to get rid of the dependence of evolution curve on initial position.
Keywords
Image segmentation, Self-adaptation, Distance maintenance