Raw-image-based image processing and computer vision algorithms usually adopt RGB Bayer patterns. Consequently, an effective approach to make the most existing image processing and computer vision tasks is to use a preprocessing method that converts o...
Raw-image-based image processing and computer vision algorithms usually adopt RGB Bayer patterns. Consequently, an effective approach to make the most existing image processing and computer vision tasks is to use a preprocessing method that converts other color filter arrays (CFAs) to RGB Bayer patterns. In particular, deep learning-based methods should be used to ensure the arrangement structure and image quality of an RGBW CFA. In this paper, we present a data preprocessing method that can produce high-quality RGB Bayer patterns. Extensive experimental results show that the proposed method outperforms conventional U-Net-based methods in terms of PSNR values.