Automatic Extraction Of The Target Area Based On Graph Cut And Its Applications In Image Retrieval

Among the various content based image retrieval methods, there is a promising method named the image retrieval based on object region. The main idea of this method is to use the feature of object region instead of the feature of global image to participate in the image retrieval. So it can be seen that the key is how to extract the object region as precisely as possible. The problem of object region extraction belongs to image segmentation. In this paper we introduce an image segmentation method based on graph cut to extract object region.Graph cut method is based on the theory of graphics, and it is one of hot spots in international image segmentation field recent years. The process of this method is that label object and background seeds by uses firstly, then extract the object region according to the feature of the labeled seeds and their adjacent pixels. So it is an interactive image segmentation method. But it is impossible that users label every image in image retrieval system, therefore, we should study a method which can extract object region automatically. As long as implying seeds label automatically, we may imply object region extraction automatically.We should train a lot of images and get their label record to complete seeds label automatically. In our training process we note down the proportion of location of labeled seeds in an image. At last we get the average of all records as the finally record. We read the seeds records and imply seeds label automatically, and then complete object region extraction automatically.In the end, we apply the object region extracted automatically into image retrieval based on object region. Take the object region as sub-image and extract its color feature, then match the object image and retrieval image according to similarity rules, at last complete image retrieval based on object region.