Technology Overview
AI-ISP Image Restoration and Enhancement Technology.
Parameter tuning of an image signal processor (ISP) is complex and usually needs considerable efforts from experienced experts. ITRI thereby proposed a novel neural network architecture Channel Spatial Attention Network (CSANet) and developed the AI-ISP Image Restoration and Enhancement Technology.
The innovation features high inference speed and peak signal-to-noise ratio (PSNR). With an end-to-end approach and deep learning algorithms, a series of ISP tasks can be completed in a short time with high image quality. The proposed method was presented in the conference of Computer Vision and Pattern Recognition (CVPR) 2021 and achieved second place in the Learned Smartphone ISP Challenge at the Mobile AI Workshop.