MMM: a stochastic mechanism for image database queries

摘要: We present a mechanism called the Markov model mediator (MMM) to facilitate effective retrieval for content-based image (CBIR). Different from common methods in retrieval, our stochastic not only takes into consideration low-level content features, but also learns high-level concepts set of training data, such as access frequencies and patterns images. The advantage proposed is that it exploits structured description visual contents well relative affinity measurements among Consequently, provides capability bridge gap between features concepts. Our experimental results demonstrate MMM can effectively assist retrieving more accurate user queries.