Objective Evaluation System for 2K Video Quality Transcoded from 4K Source using Dynamic Thresholds and Statistical Methods

Objective evaluation methodologies can be applied only to comparing images of the same resolutions, leaving several problems in assessing video quality across different resolutions. Our proposed method employs a spatial two-dimensional filter to enable pixel-by-pixel comparisons for automating the evaluation of 2K video quality transcoded from 4K. Detection thresholds are adapted for spatially and temporally complex videos, minimizing false alarms. For noise detection in specific regions, a small-area PSNR metric is applied with a statistical approach for enhanced accuracy. This automated approach matches or exceeds human review accuracy, offering cost reduction and improved quality assurance in 4K/2K productions.

Nariaki Takahashi | Japan Broadcasting Corporation | Shibuya, Tokyo, Japan
Keita Kataoka | Japan Broadcasting Corporation | Shibuya, Tokyo, Japan