Reducing the effects of quantization error in image compression systems: estimation of wavelet reconstruction filters using the LMS algorithm

The results of an investigation into the use of the least mean square (LMS) algorithm as a tool for the estimating the proper reconstruction filters in a wavelet-based compression system are reported. The LMS algorithm is used to determine the inverse model of the process responsible for creating the correlated noise. The LMS coefficients describing this inverse model can be stored along with the compressed data. Upon reconstruction, these coefficients can be preconvolved with the wavelet filter pair, g(n) and h(n), creating a new filter pair, g´(n ) and h´(n). When this new filter pair is used to reconstruct the compressed signal, the effects of distortions are reduced