Advanced fault-tolerance techniques for a color digital camera-on-a-chip

Color digital imagers contain red, green and blue subpixels within each color pixel. Defects that develop either at fabrication time or due to environmentally induced errors over time can cause a single color subpixel (e.g., R) to fail, while leaving the remaining colors intact. This paper investigates seven software correction algorithms that interpolate the color of a pixel based on its nearest neighbors. Using several measurements of color error, all seven methods were investigated for a large number of digital images. Interpolations using only information from the single failed color (e.g., R) in the neighbors gave the poorest results. Those using all color measurements and a quadratic interpolation formula, combined with the remaining subpixel colors (e.g., G and B) produced significantly better results. A formula developed using the CIE color coordinates of tristimulus values (X, Y, Z) yielded the best results.