Key Technique Research Of Images Vertical Search Engine

With the improving performance of internet bandwidth, software and hardware, the requirement of multimedia data searching is increasing also. Since image retrieval is the first step to cross from text retrieval to multimedia retieval, it has already been a hot research field. This paper focuses on image retrieval field and researches both context based Image Retrieval and content based image retrieval.Firstly, a web spider of context image retrieval from internet is studied to grab web page and defining the domain topic. DOM tree and Regular Expression is used to extract context, then the context is segmentted and indexed. At the same time, a mobile module is added in this search engine to propose an implementation method of mobile search engine, and a knowledge-based tree-like representation is introduced to present individually recommend.Secondly, content based image retrieval is researched. Image is divided into the multi-level regions. For each of them, the combined feature is extracted and inputted into the trained Fuzzy SVMs to classify, which has been proved better than conventional SVMs in the generalization ability. The values of classification in each category are calculated. Based on these values, the keywords are assigned.At last, visble features of image texture and color are extracted as indexes to catch the content characters of images based on the retrieval. They are used to transfer color from a source image to a target image, which builds a pixels relationship between the two images to render color. For one thing, the texture feature is used to compute the similarity between colorful image and grayscale image to search a similar source image which can help to increase the successful possibility of rendering. For another, the color feature is extracted to use to select sample pixels from source image to improve the rendering effect.