This paper discusses reversible color transforms, in which a raw camera image captured using a color filter array (CFA) is from red, green, and blue (RGB) color space to YDgCoCg color space for lossless compression. We found that conventional reversible color transforms (YDgCoCg transforms) are composed of three Haar wavelets, which are simple type of wavelet. Since the finding means that the YDgCoCg transforms on the basis of other wavelets that can more accurately predict pixels of interest in the predict steps have the potential to generate more sparse signals, we replaced Haar wavelets in the YDgCoCg transforms with Cohen-Daubechies-Feauveau (CDF) 5/3 and 9/7 wavelets, which were customized on the basis of the original pixel positions in two-dimensional (2D) space. The experimental results show that the extended YDgCoCg (YDgCoCg-X) transforms, which achieve higher sparsity, outperformed the conventional transforms when they are applied to the color transform parts in the JPEG 2000 and JPEG extended range (XR) lossless modes.