A MULTIPLE DISTANCE MEASURE FOR IMAGE CONTENT QUERY

Retrieving images from large and varied collections using image content as a key is a challenging and important problem. An image querying system is described using grid distance and region based distance measures by image segmentation, and a new data structure representation is presented for weighted distance metric. By predefining basic distance measure, specifying the relative importance of the region feature, and using "union", "intersect", and "difference" （or "not"） operations, more complex measure is formed, making it more flexible for the user to query images. Finally, the retrieving results using compound distance measure are given.