Leopard Image Retrieval using Region of Interest and Texture Feature
Jing Zhang, Seok-Wun Ha · 2006
In content based image retrieval system, texture based natural image retrieval can not extract texture feature efficiently and accurately, because the nature image is very complex and most of method extract feature use the whole nature image. In this paper, a new method is proposed for leopard texture image retrieval. Usually, the habitat of leopard is grass or tree, there is large difference between leopard image background and foreground, so it is easy to separate the leopard region from background using Canny edge detection, and then extract region of interest of foreground. The similarity of two images will be determined by the similarities between pairs of region of interest. The experimental results on the leopard texture image database demonstrate that the proposed method is encouraging with an average successful retrieval rate of 95.66%.