Improving Energy Consumption and Performance of IoT Based Camera Images Using Super-Resolution CNNs and Coarse-to-Fine Approach

Unfathomable convolution neural networks (CNNs) have proven popular in visual super-resolution (SR). On the other hand, deep CNNs for SR usually endure from training instability, resulting in poor image SR performance. The problem can be efficiently solved by gathering additional contextual information. The authors propose a coarse-to-fine SR CNN (CFSRCNN) to recuperate high-resolution (HR) image from a low-resolution version. On benchmark datasets, extensive experiments have demonstrated that our CFSRCNN model outperforms state-of-the-art SR methods in terms of efficiency and performance.