Image compression based on BTC-DPCM and its data-driven parallel implementation

This paper proposes an adaptive image compression algorithm suitable for highly-parallel software implementation. The proposed algorithm is based upon the block truncation coding (BTC), which is very simple but the image quality is not so good. In order to derive a better trade-off between reconstructed quality and computational complexity, an adaptive BTC algorithm coupled with the differential pulse code modulation (DPCM) is introduced. Furthermore, its highly-parallel data-driven scalable software implementation is discussed to realize an embedded multimedia communication system on a single chip for future various consumer appliances. Experimental evaluations of the proposed algorithm were performed using various benchmark images. The results illustrate that the proposed algorithm achieves over 3 dB better than the absolute moment BTC (AMBTC) with the same bit rate. In fact, evaluation result of the proposed data-driven implementation shows about 60 VGA images per second.