This technology is a Bayer pattern color image restoration device that calculates weights in six directions—north, south, east, west, diagonal, and anti-diagonal—using an advanced non-local means filter that reflects patch distances. It reconstructs green pixels by combining directional estimates and refines them using color difference information.
Most digital cameras use a single sensor and color filter array to reduce costs, meaning each pixel captures only one color, requiring the rest to be interpolated via demosaicing. Conventional demosaicing techniques often rely on pixel-level similarity, leading to inaccurate weights and difficulty in balancing weight accuracy with computational complexity.
By calculating adaptive weights based on both patch similarity and spatial local distance rather than just pixels, this technology improves both objective metrics—such as CPSNR, S-CIELAB, and FSIM—and subjective image quality, while reducing processing time. It can be applied to image signal processors (ISPs) in smartphone cameras, CCTV, and digital cameras to quickly produce sharp, full-color images from a single sensor.
N/A
