Image Retrieval using Canopy and Improved K mean Clustering

In a typical content based image retrieval (CBIR) system, target images are sorted by feature similarities with respect to the query. These methods fail to capture similarities among target images and user feedback. To overcome this problem existing methods combine relevance feedback and clustering. But clustering requires more number of expensive distance calculations. To remedy this problem we propose a new technique that combine canopy method, relevance feedback and improved k mean clustering. Canopy method reduces expensive distance calculation by measuring exact distances between points that occur in a common canopy. Improved k mean clustering automatically compute number of cluster and uses max min distance to reduce computational complexity. Relevance feedback captures exact user interest. The experiments show that our method is highly effective for image retrieval..