The conventional hybrid method of block truncation coding (BTC) and differential pulse code modulation (DPCM), namely the DPCM-BTC method, offers better rate-distortion performance than the standard BTC. However, the quantization error in the hybrid method is easily increased for large block sizes due to the use of two representative levels in BTC. In this paper, we first derive a bivariate quadratic function representing the mean squared error (MSE) between the original block and the block reconstructed in the DPCM framework. The near-optimal representatives obtained by quantizing the minimum of the derived function can prevent the rapid increase of the quantization error. Experimental results show that the proposed method improves peak signal-to-noise ratio performance by up to 2dB at 1.5 bit/pixel (bpp) and by 1.2dB even at a low bit rate of 1.1 bpp as compared with the DPCM-BTC method without optimization. Even with the additional computation for the quantizer optimization, the computational complexity of the proposed method is still much lower than those of transform-based compression techniques.
- block truncation coding
- differential pulse code modulation
- quantizer optimization
ASJC Scopus subject areas
- Atomic and Molecular Physics, and Optics