Combined Demosaicing and Adaptive Filtering on CFA Images

The gate-count of Image Signal Processing (ISP) units in digital cameras is significantly dominated by the amount of line buffers utilized by the CFA (Color Filter Array) image interpolation algorithm. Subsequent image filtering on RGB data requires further line buffers for each color channel. We present a local-adaptive Wiener filter for noise cancellation and image enhancement with incorporation of Chang's Variable Number of Gradients interpolation method deploying a 5×5 neighborhood for each pixel. Four line buffers in total are utilized for demosaicing and filtering. Our experimental results indicate a significant increase in SNR whereas edges are preserved and image contrast is enhanced.