Application of High-Frequency Redecomposition to Remote Sensing Image Compression

This paper proposes a new image compression scheme based on high-frequency bands redecompression according to the characteristics of remote sensing images which contain man-made objects. By the scheme, the image is decomposed at first using a symmetric-antisymmetric multiwavelets. LH subband and HL subband are decomposed in horizontal, vertical direction individually while LL subband is decomposed iteratively. This decomposition approach achieves better high-frequency distinguishability without increasing computational complexity too much. Then, SPIHT is used to compress transformed coefficients. Each coefficient at a given subband is related to a set of coefficients at the next finer subband to create parent-children relationship and it is a better method exploiting interscale dependencies. Experimental result confirms the efficiency of the new scheme to remote sensing image.