This technology is a demosaicing method that interpolates missing pixels by separating green, red, and blue channels from color filter array image sensor data and using inter-channel color differences as weighting factors. It then refines the combined plane based on interpolation directions determined by spectral correlation to produce the final color image.
Using three image sensors requires additional optical splitters and electronic controllers, which increases camera size and cost. Therefore, most cameras use a single sensor with a Bayer filter, which requires estimating the remaining channels since only one color is captured per pixel. In particular, inaccurate reconstruction of the green channel, which contains the most information, leads to a degradation in overall color image quality.
This technology calculates four-directional correlations for the green channel of the target pixel and estimates missing green components using neighboring G-R and G-B color differences and absolute color gradient-based weights. It then re-estimates the four directions using red-green color differences and uses the inverse of the gradient as a weighting factor to precisely refine the result using one of three interpolation formulas: horizontal, vertical, or normalized. It can be applied to single-sensor image chips and image reconstruction circuits in low-cost camera modules, improving G-channel accuracy and enhancing image quality relative to cost without additional optical components.
N/A
