Superpixels Saliency Detection Based On Regions

With the rapid development of science and technology, people also have an increasinglyhigh requirements for visual perception. The large amount of digital image information needsimage saliency detection to improve processing efficiency Saliency detection in digital imageprocessing has become an indispensable tool. The research of image saliency detection involvesmany relative algorithms, most of which are based on pixels, fixed block or regions.This paper presents an algorithm, and that is superpixels saliency detection based on regions.This algorithm takes the advantage of that saliency detection based on regions makes processingfaster and easier, and it is also combined with the characteristic of that saliency detection basedon pixel can obtain better saliency map. Finally it mainly improves the shortcoming of thatsaliency detection based on the region is not so precise.The article mainly includes four aspects as followed:1. Introduced the developments of saliency detection as of today2. Elaborated the basic concepts of the saliency research closely related to the human visualsystem, the two models of visual attention, saliency detection.3. Presented several typical researching saliency algorithms in detail, and analyzed thesealgorithms.4. The algorithm firstly used a SLIC superpixels segmentation method to divide the areas ofimage, then found the area of contrast, up-sampling to make up for the contrast of pixel level.After that the saliency of whole image was obtained combined with the regional location impactfactors. We took the steps of the algorithm to the different areas that had been divided from theimage, then calculated the average value by superimposing the single saliency map, finally gotthe saliency map of algorithm. What’s more, we carried out the simulation experiment an drewthe Precision-recall curve. The result turned out to be good and we got the expected results.