Research on the Image Semantic Understanding Pattern based on the Sparse Coding and Wavelet Theory

Abstract

Abstract. In this paper, we conduct research on the image semantic understanding pattern based on the sparse coding and wavelet theory. From the perspective of image understanding, according to the different weights of spatial information to give false points pixels, make the false points of pixels in different positions have different importance. Was introduced to the weighted formula of relative distance, foreground and background search distance, in order to obtain evaluation algorithm for image segmentation is scaling invariance conclusion in order to overcome large ratio of segmentation quality evaluation distortion phenomenon, put forward a kind of distortion of punishment, improve the effectiveness of the evaluation algorithm and global. Through applying the algorithm into primary application scenarios, we could test the effectiveness and robustness.