Weak supervised image semantic segmentation method based on superpixels and conditional random field

Abstract

The invention discloses a weak supervised image semantic segmentation method based on superpixels and a conditional random field, and belongs to the field of image processing. The method is characterized in that the number K of pre-segmented super-pixels can be adaptively generated when an SLIC algorithm is used for segmenting an image, color features and texture features are fused for super-pixelcombination, and the iteration termination condition is that the number of combined super-pixel blocks is equal to three times of the number of category labels. An undirected graph model is constructed by taking the super-pixel blocks as nodes; the category association information and the similarity in the image are added into a pair of potential energy functions; the super-pixel blocks are enabled to be in one-to-one correspondence with the category labels, a second-order conditional random field energy function is taken as a semantic segmentation model to carry out label inference, and an inference result is a result of optimizing the energy function, and the target classification is converted into a problem of minimizing the energy function. According to the method, a superpixel segmentation algorithm is improved, and a conditional random field model is introduced, so that the precision of weak supervised semantic segmentation is improved.

Description

technical field [0001] The invention belongs to the field of image processing, in particular to a weakly supervised image semantic segmentation method based on (Superpixel) and conditional random fields (Conditional Random Fields). Background technique [0002] Among the human perception systems, the visual system has become one of the most commonly used ways for us to obtain external information because of its large amount of information and high utilization rate. How to simulate the process of human beings from receiving a picture to semantic interpretation is a huge challenge for computer vision technology today. Nowadays, intelligent systems related to computer vision are used in every corner of social life. As an important field of computer vision research, semantic segmentation is of great significance to many applications such as scene understanding, object recognition, image or video editing. Compared with ordinary image classification, image semantic segmentation c...

AI Technical Summary

Problems solved by technology

[0010] In order to solve the problem that it is difficult to obtain pixel-level annotations for fully supervised image semantic segmentation, and the accuracy of traditional weakly supervised image semantic segmentation is not high, this paper proposes a weakly supervised image semantic segmentation method based on superpixels and conditional random fields.

Images