CNN Based Multi-view Image Quality Enhancement
Gyu-Lee Jeon, Hee-Jae Kim, Eun Yeo, Je‐Won Kang · 2022 Thirteenth International Conference on Ubiquitous and Future Networks (ICUFN) · 2022
We propose a CNN-based multi-view image quality enhancement (MVIQE) to improve the quality of a target image using adjacent multi-view images. Differing from the conventional single frame quality enhancement, our approach aims to improve the quality by transferring a higher quality of other input views in multi-view images as references. Our network contains an optical flow estimation module, warping layers, and image synthesis module to enhance the quality of a target image with quantization noise. Experimental results show that our method outperforms previous studies on image quality enhancement in terms of peak signal-to-noise ratio performance.