Optimized filter banks for the compression of subtraction angiographic images

Abstract

Several classes of images contain a clearly determined set of basic signals. This permits the development of signal-matched filter banks for image compression. A new approach for the derivation of signal-optimized filter banks using nonlinear optimization is presented here. Filters optimized with respect to the vessel and background signal of angiographic images are provided as a design example. Experimental results show that the new filter bank is competitive with the best biorthogonal wavelets.