A proposed video super-resolution reconstruction strategy using wavelet multi-scale convolutional neural networks

High-resolution (HR) images are often required for most applications, as they incorporate complementary information. However, the optimal utilization of sensor and visual technologies to improve image pixel density is often limited and prohibitively expensive. As a result, employing an image processing method to build an HR image from a low-resolution (LR) one is an efficient option. The goal of video super-resolution is to restore intricate points and reduce the sensor effects. This research builds on the multi-frame super-resolution approach based on wavelet analysis and convolutional neural networks (CNNs). For that purpose, the strategy begins by applying wavelet decomposition on video frames to get multi-scale representations. Then, several CNNs are trained independently to approximate wavelet multi-scale characterizations. The trained CNNs do inference by regressing wavelet multi-scale characterizations from LR frames, followed by wavelet reconstruction, which produces the reconstructed HR frames. This research presents a learning-based method for obtaining HR frames from the LR frames captured with various camera zoom lenses. The experimental findings confirm eligibility of the proposed strategy for restoring video frames.