Lightweight Stereo Image Super-Resolution Using Parallax Attention

Over the past decade, advancements in cellular phone technology have drastically transformed mobile phones from mere communication devices into powerful mini-computers. The integration of high-quality cameras with smartphones has led to a surge in demand for high-resolution images captured using these compact handheld devices. Prevalent use of dual cameras in smartphones has further motivated the exploration of stereo image super-resolution (SSR) algorithms, which utilizes the complementary information provided by binocular system. In this paper, we propose a novel and simple SSR approach that incorporates depth-wise separable convolutions and a parallax attention module. The algorithm effectively employs depth-wise separable convolutions to extract features from both left and right view of stereo image pair. Additionally, the parallax attention module facilitates the integration of cross-view information to enhance the super-resolution process. By combining these techniques, the proposed SSR model, Light weight Parallax Attention Stereo Image Super-Resolution(L-PASSR) achieves a light weight design while still delivering comparable results in stereo image enhancement.