Title :
Edge Map-Based Texture Synthesis via Graph Cuts
Author :
Zou, Kun ; Han, Guoqiang ; Li, Wen ; Zhang, Xiaoyuan ; Gu, Guosheng
Author_Institution :
South China Univ. of Technol., Guangzhou
Abstract :
This paper presents an edge map-based texture synthesis algorithm using graph cuts to find the optimal patch boundaries. It takes the whole sample texture as a patch. In each step, an optimal patch translation with minimum match error is found, and graph cut technique is used to determine the optimal patch boundaries. In the patch translation searching process, edge map is used to measure the structural similarity, and it ensures the edge continuity better preserved. In the graph cut phase, heuristics are introduced to get around the issue that the energy function is not a metric. In this way, patch boundaries are more seamless. Experimental results show that the performance is satisfactory.
Keywords :
edge detection; graph theory; image texture; edge continuity; edge map-based texture synthesis; graph cuts; optimal patch boundaries; patch translation searching process; Computer errors; Computer graphics; Computer science; Computer vision; Cost function; Image edge detection; Image processing; Image recognition; Information technology; Optimization methods;
Conference_Titel :
Convergence Information Technology, 2007. International Conference on
Conference_Location :
Gyeongju
Print_ISBN :
0-7695-3038-9
DOI :
10.1109/ICCIT.2007.86