Calidad del video digital: modelos y métricas de visión

0
323

In “Digital Video Quality: Vision Models and Metrics” by Stefan Winkler, published in 2005 by John Wiley & Sons, the author addresses the complexity and necessity of accurate vision models and quality metrics for the evaluation of digital video. This need has grown with the expansion of digital systems in television, photography, and other forms of visual processing. Traditional analog systems are still in use, but production studios and service providers have increased their adoption of digital video equipment, making the limitations of old quality assessment techniques evident. The distortions introduced by digital compression fundamentally differ from analog ones, requiring new ways to measure perceived quality beyond conventional methods such as mean squared error (MSE) or peak signal-to-noise ratio (PSNR).

The book aims to provide an introduction to vision modeling in the context of video quality assessment. Both model design and metrics are discussed, providing examples of their use. The models presented have a general application in image and video processing. Specifically, the book’s content is organized starting with an overview of the human visual system, describing its anatomy, physiology, and the resulting perceptual phenomena.

Next, the main aspects of visual quality are explored, with a focus on digital video. This includes an introduction to video coding techniques and the analysis of the effects that compression or transmission errors have on quality. Factors influencing subjective quality are examined in depth, and procedures for its measurement are described, also reviewing the history and state of the art in video quality metrics and their performance evaluation.

The book also presents tools for vision modeling and quality measurement. Among these tools is a unique isotropic local contrast measure based on analytical directional filters and a perceptual distortion metric (PDM) that evaluates video quality considering color perception, the multi-channel architecture of temporal and spatial mechanisms, contrast sensitivity, pattern masking, and channel interactions.

Additionally, a chapter is dedicated to evaluating the predictive performance of the PDM and its comparison with competing metrics, using extensive data from subjective experiments. The design choices of the different PDM components are analyzed to understand their impact on predictive performance. Subsequently, extensions to the PDM are explored for predicting perceived blocking distortions and supporting object segmentation, also integrating visual attractiveness attributes such as sharpness and color vividness.

Finally, the book concludes with a perspective on promising developments in the field of video quality assessment, suggesting that significant advancement of the state of the art will require a deeper understanding of the human visual system to design reliable models .

Review on the publisher’s website

Review on Amazon