基于匹配误差改进和迭代传输的纹理传输算法  

Texture Transfer Algorithm Based on Matching Error Improvement and Iterative Transfer

作　　者：谭永前 曾凡菊 Tan Yongqian;Zeng Fanju(School of Big Data Engineering, Kaili University, Kaili, Guizhou 556011, China;College of Optoelectronic Engineering, Chongqing University, Chongqing 400044, China)

机构地区：[1]凯里学院大数据工程学院,贵州凯里556001 [2]重庆大学光电工程学院,重庆400044

出　　处：《激光与光电子学进展》2018年第5期124-128,共5页Laser & Optoelectronics Progress

基　　金：贵州省教育厅青年科技人才成长项目[黔教合KY字(2017)335];贵州省科技合作计划[黔科合LH字(2017)7161号;黔科合LH字(2016)7318号;黔科合LH字(2015)7744号];国家自然科学基金(11464023)

摘　　要：基于样本块的纹理传输原理,研究了纹理图像的边缘梯度信息、最小误差路径以及非线性亮度/色度型(YUV)颜色空间的亮度信息对纹理传输结果的影响。在误差度量时,将纹理图像本身的边缘梯度信息和重叠区域的最小误差路径同时作为纹理合成时的度量,并引入源纹理图像和目标图像在YUV颜色空间的亮度信息作为纹理传输的约束度量,能使找到的匹配块更适合纹理传输。同时,在纹理传输过程中采用迭代的方法,解决了大部分纹理通过一次传输后的传输效果不佳的问题。实验结果表明,改进后的纹理传输算法能取得较传统纹理传输算法更好的传输效果。Based on texture transfer principle of sample blocks, the influences of edge gradient information of texture image, the minimum error path and luminance information in the nonlinearity luma/chroma （YUV） color space on texture transfer results are studied. We take edge gradient information of texture image and the minimum error path of overlapping region as the measure of the texture synthesis at the same time, and introduce luminance information of the source texture image and the target image in YUV color space as constraint measure of the texture transmission. The found matching block is more suitable for texture transfer. At the same time, we use iterative method in the process of texture transmission to solve the problem that the transmission effect is not good after the texture transmission for one time. Experimental results show that, compared with the traditional texture transfer algorithm, the improved texture transfer algorithm ean aehieve better transmission effect.

关 键 词：图像处理 纹理传输 边缘梯度 亮度信息 迭代传输 

分 类 号：TP391.41[自动化与计算机技术—计算机应用技术]