A coding method of Bayer pattern image based on Wyner-Ziv structure

A three-channel distributed codec for Bayer pattern image is proposed based on Wyner-Ziv structure in this paper. First, Bayer pattern image is separated and converted into four component images by changing its structure, and three of them are transformed into frequency domain by the discrete cosine transform. The remaining one and the second component image will be side information after being compressed according to JPEG-LS. We extended the Lloyd algorithm and designed a optimal quantizier over all the codec according to Lagrangian cost function. The discrete cosine transform coefficients quantized in each channel are encoded independently by employing Slepian-wolf coder. At the decoder, the side information is available and Bayer pattern image is reconstructed. The experiments show the proposed method can improve rate-distortion performance of Bayer pattern image codec at high rate.