Image reconstruction using a 21 M-channel perfect reconstruction filter bank with a modified synthesis part based on an optimized adaptive interpolation kernel

Abstract

In this paper, we propose a new method for reconstruction and arbitrary scaling of images by utilizing a 2D M-channcl perfect reconstruction filter bank structure with a modified synthesis part bused on an optimized adaptive interpolation kernel. In particular, the proposed approach leads to sharper reconstructed and scaled images than the direct conversion, still preserving high frequency components of the original image using the filter bank structure. Also, it can reuse directly the decomposed signal from an analysis part of a filter bank and apply the subband processing for image conversion. The simulation results show the proposed approach yields images of better quality than direct methods which use conventional interpolation kernels.