Enhanced Content Based Image Retrieval Using Machine Learning Techniques

Recently researches have shown that there is a rapid growth of multimedia resources. There is an urgent need of intelligent methods to organize and process multimedia data to make full utilization of multimedia resources. Content-based image retrieval (CBIR) is one such process which is used in retrieval of image data. It uses a low-level feature like shape, surface, dimension etc which depict the image content. But finding a solitary best representation of Image content is nearly impossible that can isolate diverse classes with very much characterized boundaries. Moreover CBIR using distance measure is not effective while considering the low-level features. Retrieval process using classification technique alone greatly depends on the performance of classifier being used and these frameworks yield high recovery precision and for every misclassification the frameworks result in complete unsuccessful results. Alone use of these strategy does not yield sufficient results. This paper proposes a novel approach to content-based image retrieval by application of genetic algorithm that optimizes the weights gained from features extractions of an image and then integrating with machine learning techniques such as clustering that handles the ambiguity, uncertainty and imprecision when accessing the multimedia data that addresses the limitations of both traditional distance based metrics and traditional classifier grounded retrieval approaches. The proposed approach reduces the overall search time of finding the image with accuracy.