Large-scale face images retrieval: A distribution coding approach

Great progress in face recognition technology has been made recently. Such advances will provide us the possibility to build a new generation of search engine: Face Google, searching from person photos. It is very challenging to find a person from a very large or extremely large database which might hold face images of millions or hundred millions of people. The indexing technology used in most commercial search engines like Google, is very efficient for text-based search, unfortunately, it is no longer useful for image search. A solution is to use partial information (signature) about all the face images for search. The retrieval speed is approximately proportional to the size of a signature image. In this paper we will study a totally new way to compress the signature images based on the observation that the face signature images and the query images are highly correlated if they are from the same individual. The face signature image can be greatly compressed (one or two orders of magnitude improvement) by use of knowledge of the query images. We can expect the new compression algorithm to speed up face search 10 to 100 times. The challenge is that query images are not available when we compress their signature image. Our approach is to transfer the face search problem into the so-called ¿Wyner-Ziv coding¿ problem, which could give the same compression efficiency even if the query images are not available until we decompress signature images. A practical compression scheme based on LDPC codes is developed to compress face signature images.