Content Based Image Retrieval System Using Color Quantization

With the development of the Internet at large scale and the availability of various image capturing devices such as digital cameras, smart mobile phones, image scanners, digital image collection is increasing rapidly. With the result of that, content based image retrieval has grown in different areas to research. Efficient image retrieval tools are needed for users in various domains including remote sensing, crime prevention, fashion, medicine, publishing, architecture, etc. To achieve content based image retrieval, many retrieval systems have been developed. Two frameworks are accessible to perform the information retrieval which are text based retrieval and content based retrieval. But text based approach has disadvantages, firstly, more manual work is required to annotate the images and second is annotation inaccuracy. Therefore, to overcome these drawbacks, content based retrieval becomes a major topic in research. In this work, we have chosen this topic for our thesis work and explored autonomous sub-class based image retrieval depending upon color and texture of the query image dominant object by using uniform quantization and fuzzy methods of clustering to reduce the colors in the database images. Texture colors of the object have been used as a variable parameter to retrieve images according to human perception vision. The experimental results show efficient outputs in terms o recall and precision ratio.