A computational kernel for fast and efficient compressed-domain calculations of wavelet subband energies

This paper develops algorithms for calculating commonly used energy measures directly from the data stream of waveletcompressed images. Incorporating these techniques into an underlying computational kernel allows many image-processing tasks to be efficiently implemented in the compressed domain. Compared to traditional decompress-process methods, the proposed techniques offer significant memory savings and reduce the computational load of the system. Experimental results show that the calculated energy values are accurate and obtained with less strain on system resources. The potential savings achieved by incorporating the proposed kernel are illustrated in a texture classification example.