Image Saliency Region Detection Based On Convex Hull

The detection of saliency regions has become one of the important research topics in recent years.The purpose is to make the computer imitate the human visual attention mechanism to extract the regions where the target may appear from the complex scene,so as to allocate computer resources reasonably in the subsequent processing links,so that the resources can be Make full use to improve operating efficiency.Due to the rich and complex image content,it is very challenging to detect salient areas with high precision.Existing algorithms based on convex hull extraction of salient regions often contain more background noise in the convex hull,and the accuracy of the algorithm results is not ideal.In order to improve the accuracy of detection,this paper proposes three saliency detection algorithms based on convex hulls.The first method exponentially fuses the saliency map calculated by using the color and spatial position difference between the inside and outside of the convex hull and the saliency map based on the background prior obtained by using the image boundary information.The obtained result shows that the background noise can be better suppressed.In order to further improve the accuracy of the bayesian model algorithm for the detection of salient objects,the second method uses manifold ranking algorithms to extract the foreground of the image,as the prior probability of the bayesian model,and calculate the three color spaces of RGB,HSV and CIELab.The intersection of the convex hulls below is used as the minimum convex hull,and the color histogram is combined to calculate the observation likelihood probability.Finally,the bayesian model is used to calculate the saliency map,which improves the boundary definition of the saliency map.Considering that the prior map using a single algorithm cannot adapt to various complex environments,the third method makes full use of the complementarity of the two algorithms,and obtains the saliency map based on the convex hull prior and the background prior and the manifold ranking algorithm.The exponential fusion of the prior map is used as the prior probability in the Bayesian model,and the saliency map obtained by combining the improved Bayesian model has higher accuracy.The three proposed algorithms are evaluated in stages on the two public databases of MSRA and ECSSD,and compared with the traditional classic algorithms.The results show that the F-measure values of the three algorithms are improved compared with the classic algorithms.The extracted saliency map better highlights the content theme in the background blur application experiment,and the application experiment in the content scaling technology retains the structural characteristics of the saliency region,which further proves that the saliency detection algorithm proposed in this paper is practical in application value.