An enhanced dynamic multi-grouping approach for AMBTC images using pairwise grouping

Abstract Absolute Moment Block Truncation Coding (AMBTC) is a cost-effective image compression technique that replaces pixel blocks with two quantization values and a binary map to minimize storage requirements. However, it is less effective for compressing flat and complex blocks. To address this issue, a Dynamic Multi-Grouping (DMG)-based AMBTC method was proposed, which removes the binary map for flat blocks and uses ternary or quaternary maps for complex blocks. Indicators are used to track the encoding methods applied to each block. While this method improves image quality, it fails to account for the correlations among indicators and maps, leading to higher storage requirements. This paper presents a sub-grouping strategy that groups indicators with ternary or quaternary maps to exploit their correlations, and encodes them using entropy coding to reduce the encoding length. During decoding, the compressed codes of indicators can be extracted from the bitstream, and the decompressed indicators are used to reconstruct blocks of various types. Our approach can be integrated with existing DMG methods to reduce bitrate while preserving image quality. When tested on 100 images, it achieved average reductions in bitrate of 31.83% for indicators, 14.68% for ternary maps, and 8.29% for quaternary maps.