Research On Salience Algorithm Based On Image Edge Information Combined With Spatial Weight

Significance detection is mainly based on psychology and computer science and other disciplines of theoretical knowledge,the use of computer simulation of human vision system for rapid image information processing.The main application of saliency detection is to provide accurate preprocessing methods for image segmentation,image compression and other image processing techniques,in order to improve the overall processing efficiency of the algorithm.In this paper,based on the analysis of the shortcomings of the current mainstream algorithms,combined with the characteristics of human vision,two significant region detection algorithms are proposed.In order to solve the problems of more background noise,insufficient brightness and integrity of the significant region in the current mainstream saliency detection algorithms,a saliency detection algorithm combining edge information and spatial weight calculation is proposed.Firstly,the smoothed image is segmented by super-pixel.and the super-pixel segmentation as the preprocessing process of the image can greatly improve the computational efficiency.The color and brightness differences between the image and its edge region are used to calculate the preliminary significant map,and then the final significance detection results are obtained by calculating the spatial weight of the position information.Considering the limitations of the third chapter algorithm,we improve it on the basis of the third chapter algorithm.In this paper,a significant detection algorithm based on the difference between edge information and foreground information is proposed.On the basis of generating the preliminary salient map,the foreground prior region is determined by combining the convex hull region,the difference between the image and the foreground prior region is calculated,the foreground prior salient map is generated,and the two parts of the salient map are fused.Then with the help of the idea of spatial weight and S-shaped curve optimization,the final significance detection results are obtained.At the same time,this algorithm is compared with other algorithms in subjective,objective and other evaluation indicators.The experimental results show that the two algorithms can obtain more accurate significance region,less background noise and more highlight the target region.In the objective evaluation index PR curve,ROC curve,F-Measure value,MAE value are higher than other classical algorithms,The effectivenss of the proposed algorithm is proved.