Resumen :
Esta tesis aborda el desarrollo de técnicas avanzadas de codificación perceptual aplicadas al estándar de codificación de vídeo HEVC (H.265). En un contexto de creciente demanda de contenido audiovisual en alta resolución, resulta crucial desarrollar algoritmos de compresión más eficientes que mant... Ver más
This thesis addresses the development of advanced perceptual coding techniques applied to the HEVC (H.265) video coding standard. In the context of increasing demand for high-resolution audiovisual content, it is crucial to develop more efficient compression algorithms that maintain the visual quality perceived by the viewer.
In the first section, the performance of various coding tools present in the HEVC standard is investigated in depth from a perceptual perspective. Using objective metrics such as SSIM, MS-SSIM, and PSNR-HVS-M, different coding techniques are compared, evaluating both their impact on subjective visual quality and their efficiency in compression. The study reveals important findings on how to optimize the rate-distortion (R/D), laying the groundwork for future improvements in perceptual coding.
The second section of the thesis introduces a methodology that combines contrast sensitivity models and texture masking to dynamically adjust the quantization parameters (QP) based on visual content. This hybrid approach allows the encoder to better adapt to scene characteristics, maximizing compression efficiency in regions of higher perceptual relevance, preserving the most important visual details for the viewer.
Finally, in the conclusions, the main contributions of the thesis are summarized, highlighting the advances in perceptual coding and their implications for the future of video standards. Additionally, future development lines are proposed, focusing on real-time implementation and the generalization of the presented models to other video compression standards.
Overall, the contributions of this thesis represent significant advances in perceptual video coding, highlighting improvements in compression efficiency without compromising visual quality. Furthermore, future development paths are proposed, focusing on real-time implementation, the use of neural networks, and the generalization of the models presented to more recent video compression standards.
|