Image salient region detection method based on background seed points

Abstract

The invention discloses an image salient region detection method based on background seed points. The method comprises the following steps: (1) performing weighted median filtering processing and Canny edge detection processing on an input image; (2) performing super-pixel segmentation on the image obtained by weighted median filtering to obtain a super-pixel image; (3) according to the superpixelimage, taking the image gradient spectrum as an energy spectrum, finding a plurality of paths formed by superpixel regions based on a minimum energy loss path algorithm, and taking the superpixel regions through which the paths pass as background seed points; (4) performing calculating based on the background seed points to obtain a salient value of the salient region of the image; (5) fusing thesaliency values of the paths to obtain an initial saliency map; and (6) constructing foreground seed points based on the initial saliency map, and performing calculating to obtain a final saliency map. According to the method, the problem that the generated saliency region is incomplete due to the fact that the foreground region is falsely detected as the background seed point in a traditional method can be solved, and the saliency map obtained through detection is more complete and uniform.

Description

technical field [0001] The invention belongs to the technical field of image processing, and in particular relates to an image salient region detection method. Background technique [0002] Content-based image scene recognition algorithms have received extensive attention in recent years. With the increasing popularity of image capture devices, the application range of image-based scene recognition methods is gradually expanding, including medical care, surveillance, intelligent transportation, video structuring, etc. It has wide application value. As we all know, the human visual system can always capture some salient features when observing things, which can help us make a good judgment. This inspires that for the scene recognition algorithm, obtaining the visual salient area of ​​the image is very important for the scene recognition process. It will also be very beneficial. In recent years, many methods of saliency detection have been proposed at home and abroad, in whi...

AI Technical Summary

Problems solved by technology

In recent years, many methods of saliency detection have been proposed at home and abroad, in which image edges are used as strong background priors, these regions are used as background seed points, and then the saliency of the image is obtained by calculating the difference between other regions in the image and the background seed points The region method is widely used, but this kind of method has a disadvantage that it directly calculates the surrounding edge part of the image as the background, and when the salient object is at the edge part of the image, the seed point belonging to the foreground will be misdetected is the background seed point, the saliency region obtained in this way is incomplete and the accuracy is not enough

Images