An efficient lossless compression algorithm for Bayer pattern images

In this paper, we propose an effective sequential lossless compression scheme which performs sequentially pixel by pixel in raster-scan order. The proposed method has three major components: a causal interpolation module, a fixed predictor and context based adaptive arithmetic coder. The causal interpolation module estimates the values of past pixels in the context which are the same color as current predicted pixel. The predictor which is based on IF-THEN structure exploits structural redundancies between mosaic-like color components, and its residuals are well-modeled by a two-side geometry distribution centered at zero. An advantage of the proposed compression method is the clear separation between prediction and entropy coding of prediction errors. Finally, an effective and efficient arithmetic coder can be easily interfaced with our predictor. Experimental results show that our method outperforms all other evaluated schemes in all testing images in term of bit-rate while maintains modest complexity level.