基于矢量量化的贝尔原始数据压缩算法

The Bayer Raw Data Compression Based on Vector Quantization

提出了直接对贝尔原始数据进行矢量量化压缩的算法。首先把贝尔CFA图像从RGB色空间转换到YCrCb色空间,然后用分块法对图像进行矢量块分割,最后用训练好的码书对每一矢量块编码,实现对贝尔原始数据的压缩。采用改进的LBG算法训练码书,并对得到的码书用合并法进一步优化,最终得到性能优良的码书。实验结果表明,笔者提出的压缩算法带来的失真很小,压缩的比特率很低,且插值后恢复的全彩图像在主观视觉和客观质量评价中都能获得令人满意的效果。

This paper proposed a vector quantization compression algorithm for Bayer raw data.This algorithm firstly converted color space from RGB to YCrCb,then divided image data as vector using divided block method,finally encoded it with the trained codebook for each vector block and realized compression for raw data.This paper trains codebook using a improved LBG algorithm and optimizes codebook through combination method to acquire perfect.Experiment results show: this direct compression method acquired the lowe...