| 摘要 |
Red, green and blue pixels of a digitised image are combined to form a single vector that points to three lookup tables translating each pixel to linear Y, I, Q values thereby forming a linear image. The colour value of each pixel in the linear image, comprised of Y, I, Q "aim" values, is then sequentially processed to attempt a match to the closest available colours. These three aim values point to respective lookup tables which translate them to three corresponding intermediate numbers which, when summed, point to a larger table mapping them to a number corresponding to the closest colour in the predetermined palette. Three lookup tables translate the map number back to the linear space with corresponding mapped Y, I, Q values. The difference between these values and prior aim values constitutes error arising from the finite palette. The error is corrected with opposite error of adjoining pixels by error diffusion techniques. In one embodiment, the error alters the linear Y, I, Q values of yet unmapped pixels prior to their being converted to aim values. The linear image is precalculated so that error-correcting addition or subtraction is made prior to the pixel's conversion. In another embodiment, the value to add or subtract is stored in an array so that it may be added when the pixel is converted. <IMAGE> |
