Research On CFA Demosaicing And Image Compression Based On Bayer Color Filter Array

With the wide application of digital imaging equipment in the industrial control, consumerand other fields, the research of theory of technology-related and application is promoted. Themajority of camera equipment use single CCD based on Bayer pattern on account ofconstraints of the size, power consumption, cost, manufacturing process and other factors.Direct compression of Bayer image can avoid complicated interpolation algorithm, savehardware resources, and reduce power, which become the hot spot of research. The currentcompression of Bayer image is based on structure transform, the algorithm not only iscomplicated, but also its compression rate can hardly meet practical demands.Demosaicking algorithm is the key technology of digital cameras, and is also responsiblefor the largest cost of calculation in the imaging process. The performance of demosaickingdetermines the quality of image。That is an important two-dimension signal.Because of thehuge data it contains, image must be compressed when it is stroed or transported. In addition,the paper also presents a high-quality image interpolation algorithm for Bell. And presents anapproach to Bayer image compression based on wavelet transform.First of all, through a green component only by the adjacent decision, the direction isdefined as a binary flag, and in accordance with flag "0","1" to determine, given the differentcomponents related to neighborhood color and weight coefficients, from Here are the averagePSNR of up to37.7dB of green component reconstructed image. In the analysis of the colorcomponents based on the frequency dependence proposed has been reconstructed usinghigh-frequency part of the replacement of the green component aliasing distortion visible red(blue) components of the high-frequency sub-regional approach to reconstruct the averagePSNR large at36dB red (blue) component images. In the process of restoration of image, if the weights of different sub-band images, aswell as different magnitude sub-band coefficients are different, the quantitative thresholdrelies on each subband sample standard deviation: the higher weight chooses the betterquantify accuracy; while the lower one only have a rough quantitative. We develop adaptivebinary arithmetic coding to compress coefficients of every subband. Lossy compression ratiocan achieve3.1:1; PSNR of recovered image is42dB.The CFA demosaicing and image compression in the paper can be widely used in imagesensor equipment, which can improve the storage efficiency.