华中科技大学学报（自然科学版）2018，Vol.46Issue(1)：75-80,6.DOI:10.13245/j.hust.180115
基于四元稀疏正则的彩色图像超分辨率重建
Color image super-resolution reconstruction based on quaternion sparse regularization
摘要
Abstract
In order to keep a relationship between three channels of a color image and effectively improve recovery quality of the reconstructed color image of edge,color and other details,a color image super-resolution reconstruction algorithm based on quaternion sparse regularization model was proposed.The algorithm adopted quatemion to represent the information between three channels of a color image,and L1/2 regularization term was used instead of L1 regularization term to construct color image super-resolution dictionary learning and reconstruction model based on quatemion sparse regularization constraint.Meanwhile,in order to better characterize the image information,the demean method was used to construct high and low resolution training samples during training the reconstructed dictionary joint dictionary.To obtain a high resolution reconstruction dictionary that matches the low resolution reconstruction dictionary,the dictionary separation training method was used to generate the high and low resolution reconstruction dictionaries,respectively.Experimental results show that compared with the existing algorithms,the proposed algorithm can improve the subjective and objective evaluation indexes of the reconstructed images.关键词
超分辨率/稀疏表示/彩色图像/四元数/L1/2正则约束Key words
super-resolution/sparse representation/color image/quaternion/L1/2regularization分类
信息技术与安全科学引用本文复制引用
徐志刚,袁飞祥,朱红蕾,许亚美..基于四元稀疏正则的彩色图像超分辨率重建[J].华中科技大学学报（自然科学版）,2018,46(1):75-80,6.基金项目
国家自然科学基金资助项目(61761028,61562058) （61761028,61562058）
兰州理工大学博士科研基金资助项目. （）
