Salient regions detection and segmentation of an image is very important for many applications like content-based
image retrieval (CBIR), object recognition, and content delivery. But in the practical environment, there may be many or
heavy level of noisy in the input image. For a better effect, a new method is proposed in this paper: denoise the input
image by using a good performance algorithm, and then detect and segment the salient regions which we are concerned
with. We compare our new method with the current method, the experimental results demonstrate that our method
performs well and reaches a good result of image saliency regions detection and segmentation.