Saliency Detection Algorithm Based on Local Linear Constraint

In order to separate a large amount of redundant information in image recognition, an image saliency detection algorithm based on local linear constraints is proposed. Based on the traditional method of reconstructing the error, the image is detected with the local constraint. The traditional method is only constructed based on the background codebook. At the same time, the foreground dictionary is constructed by single-scale segmentation, and the image reconstruction residual is calculated on our method. The error value of the image region is calculated, and the foreground and background dictionary are combined at the same time. Then we adapt the process and normalize the acquired foreground and background saliency maps. Finally we test our method on public databases such as ASD, SED1, SED2, and SOD. The experimental data shows that the proposed algorithm is effective in detecting significant targets of images on the three criteria of P-R curve, F-measure value, and MAE value.