Bayer色彩濾波陣列(CFA)已經廣泛的被應用在現今的數位色彩相機，而所拍攝到的影像我們稱之為Bayer CFA影像I^Bayer。在這之前可逆色彩轉換壓縮與解馬賽克優先壓縮是現在兩種主要壓縮Bayer CFA影像的技術，然而，研究這兩種技術效能的文獻鮮少提出，在這篇論文中我們首先介紹相關的解馬賽克優先與可逆色彩轉換，接著點出兩著技術的優缺點，根據IMAX、Kodak，Video以及FiveK資料集所獲得的Bayer CFA影像，實驗結果在JPEG顯示出可逆色彩轉換壓縮在低跟高DCT係數量化的情況下比解馬賽克優先壓縮有更好的品質位元率權衡，但解馬賽克優先在中DCT係數量化的情況下有則會有更好的品質位元率權衡。而JPEG2000則顯示出解使用不同解馬賽克方法的馬賽克優先壓縮在任何壓縮率下比可逆顏色轉換有更好的品質位元率權衡。

The Bayer color filter array (CFA) pattern has been widely used in modern digi­tal color cameras, and the captured image is called the Bayer CFA image I^Bayer. Pre­viously, the reversible color transformation based (RCT­-based) compression scheme and the demosaicking­-first (DF-based) compression scheme are the two mainstreams for com­pressing the input Bayer CFA images. However, in the literature, the thorough performance comparison between the two compression schemes is rarely investigated. In this pa­per, we first introduce the related RCT­ and DF-­based compression works. Next, we char­acterize the advantages and disadvantages between the two compression schemes. Based on the testing Bayer CFA images collected from several classical images and the Ko­dak, IMAX, Video, and FiveK datasets, the comprehensive experimental results demon­strated that for compressing Bayer CFA images, the RCT­based compression scheme has a better quality­-bitrate tradeoff than the DF­-based compression scheme in low DCT(discrete cosine transform) and high coefficient ­quantization circumstance on the Joint Photographic Experts Group (JPEG), but the DF­-based compression scheme has much better quality-­bitrate tradeoff performance in middle DCT coefficient ­quantization cir­cumstances. In JPEG2000, DF-­based compression has better quality-­bitrate tradeoff than RCT-­based compression.

[1]K. L. Chung, T. Y. Liu, and J. S. Cheng, “Novel and optimal luma modification­basedchroma downsampling for bayer color filter array images,”IEEE Open Journal ofCircuits and Systems, vol. 1, pp. 48–59, 2020.
[2]G. K. Wallace, “The jpeg still picture compression standard,”IEEE transactions onconsumer electronics, vol. 38, no. 1, pp. xviii–xxxiv, 1992.
[3]A. Skodras, C. Christopoulos, and T. Ebrahimi, “The jpeg 2000 still image compres­sion standard,”IEEE Signal processing magazine, vol. 18, no. 5, pp. 36–58, 2001.
[4]“Imax dataset.”http://www.comp.polyu.edu.hk/~cslzhang/CDM_Dataset.htm. Accessed: Aug. 2014.
[5]“Kodak dataset.”https://www.math.purdue.edu/~lucier/PHOTO_CD/BMP_IMAGES/. Accessed: Aug. 2014.
[6]“Video dataset.”ftp//140.118.175.164/CFASS.
[7]V. Bychkovsky, S. Paris, E. Chan, and F. Durand, “Learning photographic globaltonal adjustment with a database of input/output image pairs,” inCVPR2011, pp. 97–104, IEEE, 2011.
[8]Z. Wang, A. C. Bovik, H. R. Sheikh, and E. P. Simoncelli, “Image quality assessment:from error visibility to structural similarity,”IEEEtransactionsonimageprocessing,vol. 13, no. 4, pp. 600–612, 2004.
[9]S. Y. Lee and A. Ortega, “A novel approach of image compression in digital cameraswith a bayer color filter array,” inProceedings 2001 International Conference onImage Processing (Cat. No. 01CH37205), vol. 3, pp. 482–485, IEEE, 2001.
[10]T. Richter and S. Fößel, “Bayer pattern compression with jpeg xs,” in2019 IEEEInternational Conference on Image Processing (ICIP), pp. 3177–3181, IEEE, 2019.
[11]H. S. Malvar and G. J. Sullivan, “Progressive­to­lossless compression of color­filter­array images using macropixel spectral­spatial transformation,” in2012 Data Com­pression Conference, pp. 3–12, IEEE, 2012.26
[12]D. Kiku, Y. Monno, M. Tanaka, and M. Okutomi, “Residual interpolation for colorimage demosaicking,” in2013IEEEInternationalConferenceonImageProcessing,pp. 2304–2308, IEEE, 2013.
[13]L. Condat, “A generic variational approach for demosaicking from an arbitrary colorfilter array,” in2009 16th IEEE International Conference on Image Processing(ICIP), pp. 1625–1628, IEEE, 2009.
[14]C. H. Lin, K. L. Chung, and C. W. Yu, “Novel chroma subsampling strategy based onmathematical optimization for compressing mosaic videos with arbitrary rgb colorfilter arrays in h. 264/avc and hevc,”IEEE Transactions on Circuits and Systems forVideo Technology, vol. 26, no. 9, pp. 1722–1733, 2015.
[15]G. Bjontegaard, “Calculation of average psnr differences between rd­curves,”VCEG­M33, 2001.
[16]Y. Lee, K. Hirakawa, and T. Q. Nguyen, “Camera­aware multi­resolution analysisfor raw image sensor data compression,”IEEE Transactions on Image Processing,vol. 27, no. 6, pp. 2806–2817, 2018.