Compactly merged arithmetic for wavelet transforms

Abstract

A new form of merged arithmetic is presented to compute wavelet transforms for image compression. Our approach is suitable for a wavelet-specific processor, which offers high-performance for image compression with wavelet transforms. This arithmetic is a compact form of merged arithmetic that is specifically optimized for the wavelet transform by eliminating bit-products, thus reducing the size of reduction. It develops a dual merging process to segregate the positive filter coefficients from the negative ones. Furthermore, it utilizes the bitmaps of the filter coefficients, fixed for a specific wavelet filter, and offers superior performance in both speed and size. Employing pipeline techniques, this approach provides an attractive circuit for the wavelet method of image compression.