Preprocessing Methods for Improved Lossless Compression of Color Look-up Tables

摘要

Color look-up tables (CLUTs) that provide transformations between various color spaces are commonly embedded in printer firmware where they are stored in relatively expensive flash memory. As the number of look-up tables in color devices increases in size, the space requirements of storing these CLUTs also increase. In order to conserve memory and thereby reduce cost, it is desirable to compress CLUTs prior to storage and restore tables as required. We consider methods for improving the performance of existing lossless compression methods for this application through computationally simple preprocessing. The preprocessing combines predictive coding and data reordering to better exploit the redundancy in CLUT data. Two predictive coding methods are considered: (a) hierarchical differential encoding methods, which generalizes differential coding to multiple dimensions, and (b) cellular interpolative predictive coding, which refines a CLUT in a coarse to fine order using interpolative prediction. Space filling curves that preserve continuity in the multidimensional CLUT structure are utilized for reordering the residuals obtained from hierarchical differential encoding. For the cellular interpolative prediction, we reorder the data in the coarse to fine order utilized for prediction. Results indicate that the proposed preprocessing methods offer significant performance improvements in comparison with direct compression. The best performance is obtained using the cellular interpolative predictive coding and corresponding reordering with the LZMA algorithm. This method provides a compression ratio of 3.19 over our representative CLUT data set, and an improvement of 31.33% over direct LZMA compression, the latter being the best performing direct method.