Texture Contrast Based Salient Region Detection

Abstract

As one of the basic properties of image, texture undoubtedly affect the image saliency. We introduce a texture contrast based salient region detection method, which first divide an input image into several nearly uniform super pixels, then analyze the texture feature and calculate the texture differences between regions to detect salient region. In order to obtain a better saliency map, we also optimize our method with hierarchical image fusion. Our experimental results demonstrate that the obtained saliency map generated by our method is clear and have better precision and recall ratio compared with traditional methods.