Convolutional sparse-coding-based 3DV image super-resolution framework

Super-Resolution (SR) reconstruction of images has an extreme importance for vision applications. Numerous algorithms have been introduced for this purpose in recent years. This paper presents a cost-effective approach for visual quality and resolution enhancement of 3D Video (3DV) sequences. The basic idea of this approach is to employ a deep learning algorithm for SR reconstruction based on Sparse Coding (SC) applied on video sequences degraded with up-sampling, blurring, and noise. The proposed approach is compared with the state-of-the-art bicubic approach. Simulation results have revealed high quality of the obtained 3DV frames with the proposed approach with appreciated outcomes of local contrast, average gradient, Mean Square Error (MSE), edge intensity, Peak Signal-to-Noise Ratio (PSNR), Structural Similarity Index (SSIM), Feature Similarity Index (FSIM), and entropy metrics. In addition, the simulation results introduce good and appreciated histogram results that prove the performance efficiency of the proposed SR reconstruction framework.