Saliency Detection Algorithm Based On Background Priori And Multi-feature Fusion

As an important research direction in the field of computer vision,image saliency detection has gradually attracted the attention of scholars.It is usually used as a preprocessing step of an image,such as image retrieval and image compression.The main purpose of image saliency detection is to imitate human visual attention mechanism to extract visual salient regions,which attract human attention most in images.Through image saliency detection operation,a small amount of important information can be selected from a large number of visual information for further processing and analysis,thereby improving image processing efficiency.Therefore,the research of image saliency detection technology is of great significance.This thesis focuses on two main lines: background priori and multi-feature fusion.In order to separate the background from the foreground and suppress the background noise to get a more pure saliency detection image,a saliency detection algorithm based on background priori and multi-feature fusion is proposed in this thesis.The first algorithm firstly extracts the boundary of the image and divides it into super-pixels,then extracts the boundary around the image according to the limited threshold to form a set of boundaries,calculates the probability of super-pixel boundaries,gets the optimized set of boundaries,and then further obtains the pure set of background nodes.Finally,the background node set is regarded as a priori knowledge,and the relationship between all the super-pixels and background nodes is calculated according to the information of different colors.Finally,the saliency detection map based on the background priori is obtained.The second algorithm uses the global feature single-level fusion method to solve the problem of local information missing saliency being only roughly labeled,which can not effectively highlight saliency target area.This thesis uses the global and local comparison method to make use of image color features,spatial features and frequency characteristics.The feature map is normalized and superimposed to get the multi-feature fusion saliency detection map,which can accurately detect the saliency target.The third algorithm is multi feature fusion significance detection algorithm based on background priori.It mainly uses the above two algorithms for complementary fusion,and balances the significance of the fused image to optimize the fused significance detection image and effectively improve the detection accuracy.The experimental results show that the complementary fusion algorithm proposed in this thesis can effectively suppress background noise,improve the separation between the significant target and the background area,highlight the significant target area,improve the robustness,obtain more complete and reliable significant targets,and significantly improve the detection performance.