An Improved Region Contrast and Global Distribution Saliency Detection algorithm

According to the local contrast and global distribution of an image, this paper detecting salient images through bottom-up data driven. First, this paper using adaptive segmentation method divided image into non-overlapping images, improved Block and Chessboar distance from a linear combination to replace the Euclidean distance method to calculate the regional features of contrast functions, then calculate the global distribution of feature functions, finally fusion of the above features for computing saliency map. The algorithm taking into account local features and global features to get more accurate saliency map. Test our method on the international public data sets MSRA-1000, the experimental result proves that the images extracted by this method are more accurate and more clearly, while reducing the calculation time of regional characteristics, having strong noise and high texture regions resistance, and can basically ignore the complex background.