Integrating image matching and classification for multimedia retrieval on the Web

With the recent explosive growth in the volume of images on the World Wide Web, it has become increasingly difficult to search for images of interest. The classification of images helps users to access a large image collection efficiently. Classification reduces search space by filtering unrelated images. We present a technique for image classification based on color, shape and composition using the primary objects. Our experimental results show that this approach can maintain 73% of recall just by searching 24% of the whole data set. The classification results are also used to improve the users´ interaction. The retrieval candidates are categorized based on the primary objects and presented to the users. We also show how we apply such technique to assist users in navigation