Abstract: In some embodiments of the invention, when a color component (e.g. red, green or blue) is reconstructed at a pixel from the intensity values for this component at adjacent pixels, the intensities at the adjacent pixels are assigned weights. Greater weights are assigned to those pixels for which a gradient is smaller. The gradient for each adjacent pixel is indicative of a magnitude of a change of another color component in the direction of the adjacent pixel and/or a magnitude of the rate of change of the other color component in the direction of the adjacent pixel. In some embodiments, the weights are chosen to reduce the impact of noise. In some embodiments, the reconstructed value of a color component is adjusted to incorporate information on color spikes that is carried by another color component.

Abstract: Methods and apparatus are provided for performing reversible color conversion on digital image data using integer computer arithmetic. This color conversion provides an approximation of the luminance component to the luminance as perceived by the human visual system with any necessary precision (Kd) and without multiplication and division operations other than shifts. The color conversion can be used in conjunction with a lossless or lossy compression/decompression process. Other embodiments, including non-reversible and non-integer color conversion, are also provided.