A Sparse Representation Method of Conformal Mapping for Image and Video Scene Content

Abstract

The present invention provides a conformal mapping sparse expression method for image and video scene content, comprising the following steps: 1. Input the original image or video and sample in the feature space; 2. Calculate the K nearest neighbors of each sample and establish local complete adjacency Graph to calculate the distance between adjacent samples; 3. According to the conformal mapping rule, combine it with the sparse representation method to learn a dictionary with conformal properties; 4. Use this dictionary to reconstruct the original image or video. The invention maintains the angle information between adjacent samples to the greatest extent by introducing the conformal mapping rule, and obtains a dictionary with stronger expressive ability. At the same time, conformal mapping encourages adjacent samples to be reconstructed with similar dictionaries, making the dictionaries more concise and compact. It has broad application prospects in the fields of image processing, computer vision and augmented reality technology.

Description

technical field [0001] The invention relates to the technical fields of image processing, computer vision and augmented reality, in particular to a conformal mapping sparse expression method for image and video scene content. Background technique [0002] In recent years, sparse representation and dictionary learning techniques have received a lot of attention as a research hotspot, and have been widely used in image processing and computer vision, such as image super-resolution, image denoising, classification and color editing, etc. The sparse representation technique is to reconstruct the signal by linear combination of samples in the over-complete dictionary, and limit the number of reconstructed samples to achieve the sparse property. [0003] At present, many researchers are devoted to the study of sparse representation methods, and dictionaries play a very important role in sparse representation techniques. Michal Aharon et al. proposed the K-SVD dictionary learning ...

AI Technical Summary

Problems solved by technology

These dictionary learning methods add the difference between categories, but do not consider the local relationship and context information in the data space, resulting in low expression ability of the dictionary

Images