Research And Application On Image Saliency Region Detection Model

With the rapid development of computer and communication technology,image and video are increasingly becoming the main form of bearing data information.How to process and retrieval images through computer rapidly is a great challenge with the explosive growth of image resources.Saliency detection technology is an important step in the field of computer vision,such as image retrieval,image adaptive segmentation and target recognition etc.The human visual system is good at helping people to search for the region of interest in the face of complex scenes,so simulating the human visual mechanism to extract the saliency region can help to priority allocate the computing resources for significant regional in the process of image synthesis and analysis,thus improving the efficiency of computer image processing.On the basis of the research of the existing saliency detection model,it is concluded that existing models have several drawbacks,such as low accuracy,blurry contour extraction and the bad anti noise ability.The main research work of this paper is as follows:(1)Combined with the advantage of local features contrast highlighting the salient object edge and the global feature contrast highlighting the internal saliency region,this paper presents a model which uses local feature and global feature contrast to extract saliency area(LGC).Firstly,the algorithm uses SLIC(Simple Linear Iterative Clustering)segmentation method to segment the image into several compact super pixels and select the boundary area set to calculate the boundary weighting parameter for each super pixel.Then the local saliency map is obtained by calculating the local contrast of color and texture features and the global saliency map is obtained by using the global feature uniqueness and spatial distribution characteristics.Finally,a new method SP(Sum and Product)is designed to integrate the local and global saliency map to get the final saliency map.The experimental result on the Achanta database demonstrates that the proposed algorithm outperforms than other 5 visual saliency detection methods in terms of accuracy.(2)The saliency detection model proposed in this paper is applied to image interest region automatic segmentation,content sensitive image resizing and untruthfulness rendering application.The experiment contrast results show that our model presents better results than other traditional models on the application of image.