Image object saliency detection using center surround contrast
Apama Shinde, Santosh Chapaneri, Deepak Jayaswal · 2017
Human visual system can quickly and efficiently estimate salient regions in a natural images. Saliency is about limiting our focus on important and/or relevant parts of data. In this paper, a method for saliency detection is proposed to improve state-of-the-art. The input color image is first converted to LAB color space followed by color quantization. The probability of saliency is then computed by determining center and variance of colors by using histogram of quantized image. Center surround contrast is found by scale space using twin pyramids. The saliency map is thus computed by fusing the probability of saliency with center surround contrast. The proposed method is tested using MSRA-1000 dataset and the result obtained is improved relative to an existing method, both subjectively as well as quantitatively in terms of AUC.