Evaluación y Comparación de Métricas Objetivas PSNR, SSIM y LPIPS para el Análisis de Calidad de Video
Year: 2025
Type: article
Abstract: This paper presents a tool for video quality assessment that allows the selection of quality (QP), temporal (FPS), and spatial (bitrate) scalability parameters. The proposal integrates traditional metrics such as Peak Signal-to-Noise Ratio (PSNR) and Structural Similarity Index (SSIM), together with the perceptual metric Perceptual Image Patch Similarity (LPIPS), which is based on deep neural networks. To validate its effectiveness, a two-phase subjective evaluation methodology was applied. In the first phase, participants assessed videos encoded with the same scalability parameter, showing a strong correspondence between visual perception and objective metrics. In the second phase, different configurations were compared, revealing a preference for high quality and intermediate spatial scalability. Additionally, in experiments with common distortions such as blurring and noise, LPIPS achieved a sensitivity of 73.64%, outperforming PSNR and SSIM in its alignment with human perception. The main contribution of this work is a tool that combines objective and subjective evaluations, enabling a more comprehensive analysis that closely reflects human visual perception.
Source: Revista Tecnológica - ESPOL
Authors: Carlos Flores Maza, Santiago González Martínez
Institution: University of Cuenca
Language: English
Subfield: Computer Vision and Pattern Recognition
Field: Computer Science
Domain: Physical Sciences
Open Access status: diamond
OpenAlex ID: W4415201462
FWCI:
Cites:
Cited by:
Related to: