Dual-Path Residual Attention Convolution Networks for Color-Embedded-Grayscale Image
Heri Prasetyo, Abid Ammar Mahdy, Abrar Dwi Fairuz Nadhif, Taufiqurrakhman Nur Hidayat, Rudi Hartono · 2023
The reversible grayscale to color recovery is a challenging task due to its ill-posed inverse nature. The recovered color image produced by the color-embedded-greyscale process commonly does not have mesmerizing quality because of color distortion and checkerboard artifacts. To tackle this problem, we propose a Dual-Path Residual Attention Convolution Network (DPRACNet) to improve the quality of recovered color image. This method implements dual-path convolution network and residual attention as the building block. As reported in the Experiment Section, the DPRACNet provides best performance among the previous methods with Peak Signal to Noise Ratio (PSNR) and Structural Similarity Index (SSIM) is 31.62 and 0.934, respectively.