Demosaicing Method Based on Anti-Aliasing FIR Filters

Bayer Demosaicing Using Anti-Aliasing FIR Filters and Inverse Gradient Refinement

This technology is a demosaicing method that estimates missing green values from the four directions—north, west, east, and south—using a 5-tap anti-aliasing FIR filter on a single-sensor camera's Bayer block, then combines them with weights and refines the interpolated green plane through inverse gradient-weighted filtering.

Since single-sensor cameras record only one color per pixel, the remaining two colors must be estimated using the spectral correlation of neighboring pixels. Conventional directional weighted interpolation often suffered from low accuracy due to reliance on unreliable estimates, particularly in areas with irregular edges and textures where low correlation leads to significant interpolation errors.

This technology enables multiple pixels to contribute to color estimation using FIR filter-based directional color differences and direction-specific gradient weights, while refining the results with a 5-point inverse gradient filter to preserve edges and details. The same process is applied to the blue and red planes. It can be integrated into the RAW image processing pipelines of smartphone and security cameras to produce clean, full-color images with suppressed zipper artifacts along edges.

‍

‍

Key Features:

• A Bayer block generation step that creates a Bayer block from the input image prior to green plane interpolation and refinement
• A step that interpolates missing green elements in the north, west, east, and south directions using a 5-tap fixed-coefficient anti-aliasing FIR filter based on the spectral correlation between blue and green
• A step that calculates normalized interpolation results for missing green colors by applying weighting factors derived from directional gradients
• A step that refines the green plane using 5-point inverse gradient-weighted filtering and interpolates/refines the blue or red planes in the same manner

‍