EBSD-Net: Enhancing Brightness and Suppressing Degradation for Low-light Color Image using Deep Networks
Minh-Thien Duong, Min-Cheol Hong · 2022 IEEE International Conference on Consumer Electronics-Asia (ICCE-Asia) · 2022
Image enhancement in the RGB color space typically leads to undesirable artifacts, especially color distortion owing to insufficient consideration of the correlation between color channels. This paper presents enhancing brightness and suppressing degradation for low-light color image using deep networks (defined as EBSD-Net). To begin with, the input image is converted from RGB color space to the YCbCr color space, and then the three channels Y, Cb, and Cr are separated from the YCbCr image. Next, the Boosting-Net is proposed to enhance the luminance channel Y with minimized halo artifacts. In addition, the Chrominance-Net is introduced to reduce color distortion of two chroma channels Cb and Cr. After the enhancing brightness and suppressing degradation are completed, the luminance-enhanced channel and two chrominance-corrected channels are transformed back into RGB color space to obtain the enhanced image. The extensive experimental results show that the proposed network outperforms other competitive methods in qualitative and quantitative terms.