Performance Assessment of the Adaptive GoP-size extension of the Wireless SoftCast Video Scheme

Abstract

The SoftCast video transmission scheme has been proposed as a promising alternative to traditional video broadcasting systems in wireless error-prone environments. However, in its original form, the scheme may introduce annoying artifacts such as temporal quality variations or the so-called ghost effect. To eliminate this artifact as well as improving the received quality, we recently proposed an updated version of SoftCast: Adaptive GoP-size mechanism based on Content and Cut detection for SoftCast (AGCC-SoftCast) [1]. In this paper, we provide additional results and verify the validity of the algorithm considering different video content and resolution.We also analyze the behaviors of the scheme considering different type of cuts: 1) Abrupt cuts i.e., a cut that appear between two different scenes, 2) Soft cuts that may appear in a same scene and 3) Intraframe cuts that represents a shot change inside a limited portion of the frame. The performance of AGCC-SoftCast are compared to the original SoftCast scheme under different Channel Signal-to-Noise Ratios (CSNR) and Compression Ratios (CR). Traditional objective metrics are considered such as: the Peak Signal-to-Noise Ratio (PSNR), the Structural SIMilarity (SSIM) and the Multi-Scale SSIM (MS-SSIM). In addition, the recent Video Multi-method Assessment Fusion (VMAF) metric proposed by Netflix is also used. Regardless of the CSNR and CR values, results clearly highlight the importance of the AGCC algorithm in a SoftCast wireless video transmission scheme. Depending on the transmitted video content, results show that improvements up to 11.6dB in terms of PSNR and 67.45 in terms of VMAF can be obtained, especially at the cut boundaries.

DeepWiVe: Deep-Learning-Aided Wireless Video Transmission

“…In [38], the prior used for the estimation is modified to obtain better reconstruction quality (SoftCast assumes a Gaussian prior). In [39], the variability of temporal redundancies between frames is addressed by introducing adaptive GoP sizes. Although these methods have been shown to overcome the cliff-effect, they are not competitive to separation-based schemes when it comes to video quality and cannot exploit the available bandwidth, or adapt to channel and network conditions dynamically.…”