Combining contrast enhancement by reverse mapping with multiresolution VQ for browsing

Abstract

Forms a multiresolution vector quantization (MVQ) codebook for progressive transmission using principal components partitioning of a full search VQ codebook. Down-sized intermediate codewords provide fast reconstruction of coarse images appropriate for browsing satellite image databases in an X-Window environment. Browse images can be expanded in size and improved in quality progressively. Any single band of a Landsat image has very low contrast, but contrast enhancement by global histogram equalization produces a visually informative image. When VQ codebook design incorporates contrast enhancement of the codebook to produce the decoder, the VQ requires no additional operations to produce a contrast enhanced final image [Cosman et al., 1992]. The present authors adapt this technique for better performance in their multiresolution scheme by first designing the decoder codebook from a contrast enhanced training set. Using the inverse technique, reverse mapping, one then derive the encoder from the decoder. Thus the decoder is optimized for the contrast enhanced image, a necessity for expansion into the multiresolution decoder.