Abstract
In order to reduce the blocking distortion on the conventional transform coding, the original image is divided into overlapped blocks. Then hierarchical images are constructed by their blocks. In this report, we propose a hierarchical scheme of predictive transform image coding which predicts the midpoint values, transforms the residual errors in terms of the discrete sine functions, and quantizes its transformed coefficients. Further we present an adaptive method of adaptive bit allocation to improve reconstructed image. In this method, the reconstructed image is found to be improved about 1.5 [dB] in comparison with without adaptive strategy. It is also confirmed that we can obtain an efficient subjective quality of the reconstructed image with less block distortion.
