A Generalized Recursive Block Coding for Image Data Compression

概要

論文の詳細を見る
In this paper a generalized recursive block coding(GRBC)scheme is put forward with a novel non-causal predictor free from the separable assumption on the original random field and then applied to an image data compression so as to compare with the conventional recursive block coding(RBC). In the presently proposed predictor to derive the residual random fields, the constraint on the separability of the original image is completely removed in general in contrast with the conventional 2-dimensional RBC based on non-causal predictive method which eventually leads to the one-dimensional RBC strategy. In addition the resultant characteristic KL functions for the residual errors in GRBC are confirmed to be substantially reduced to the same orthogonal discrete sine functions(DSFs)as RBC, whereas the corresponding eigen values are elucidated to be not expressed in the direct product form but in a somewhat generalized form. Also a novel bit allocation method for the transformed coefficients of the residuals is argued in connection with the eigen value problem for the residual random fields. Finally, introducing an adaptive zonal coding method, the presently proposed scheme is applied to the block codings to clarify a certain advantage beyond the conventional recursive block transform coding.
1998-06-25