Research on the Classification of Image Semantic Scene

This paper mainly aims at the algorithm of scene categorization. Firstly, putting forward by using the edge information of training set image to modela probabilistic generative model, and getingt the shape prior distribution of each scene category by learning, and inferencing to get the boundary distribution image, so to finish the regular super pixel grid block. Then, clustering the local word package description data of sub region, to construct the semantic visual vocabulary SVV, combining with the regularized super pixel grid to descript image; finally, finishing the task.