Optimization of filters for subband coding of images

Summary form only given. Recent results in optimization of filter kernels for subband coding of images, for both infinite-impulse-response (IIR) and finite-impulse-response (FIR) filters, have been reviewed. Aspects of orthogonality of filter banks as well as the choice between odd order and even order filters are discussed. As an optimization function, a weighted sum of the quantized band step response error (e.g. overshoot) and the frequency response has been used. Due to the difficulty of optimizing such a target function (which tend to have a large number of local minima), a version of simulated annealing has been used for optimization. The subjective difference between two images that were coded using filters with and without consideration of the step response has been examined.<>