Demosaicing According to Edge Information Based on Frequency Band Partition

In this paper, we propose a new demosaicing method that interpolates RAW images in a Bayer pattern. We firstly apply a wavelet transform to a rotated G-channel of Bayer pattern. Then, according to the high frequency energy, Sobel operator and 2D FFT (fast Fourier transform) are performed parallelly as edge-classifiers. Missing samples are interpolated along the estimated edge direction. In both subjective and objective evaluations, the proposed demosaicing algorithm show better results than those of conventional methods.