Video Picture Quality Measurements Based on Human Vision Modeling

Picture quality measurements based solely on detecting the noise differences between the reference and test videos fail to account for the characteristics of human perception and have broad limitations as a result.

By Greg Hoffman, Tektronix

Page 1 of 7
Video/Imaging DesignWire
(9/7/2009 2:00:03 AM)

In video systems today, picture quality is under constant siege.  At every step from original production to final presentation, a multitude of devices and technology stands in the way between a quality picture and one that has noticeable impairments.  First there are encoders and decoders that handle compression to keep files size manageable. Then format converters change resolutions and aspect ratios while other devices and transmission paths add noise and delay.

Meanwhile, consumers’ quality expectations have continued to rise as analog video technology transitions to digital technology and standard definition transitions to high definition. Unlike the world of analog video systems, the industry can no longer rely solely on signal measurements and picture monitors to assess picture quality. Professionals need better tools across the board-from equipment design through to content distribution – to verify that systems or processes have not introduced impairments in video content that will affect perceived picture quality.

In particular, the industry needs a substitute for the costly, time-consuming and potentially inconsistent subjective approach to measuring video picture quality currently prevalent. One approach gaining acceptance is to give instruments the ability to perceive picture quality in the same way as the human vision system. Such an approach leads to an objective, repeatable evaluation of picture quality that is consistent with subjective picture quality evaluation.  This gives video system designers as well as content producers and distributors a powerful ally in the quest for better picture quality.

Subjective Assessment and Objective Picture Quality Measurement

Subjective testing, whether performed according to ITU-R BT.500 specifications or informally with “golden eyes,” requires a considerable commitment of time, resources and expense, and even under the best cases isn’t always repeatable. On one hand, a golden eye might spot impairments that average viewers can’t identify, leading to expense trying to fix non-relevant problems. On the other, panels of human viewers may become tired and miss significant impairments.

If they use these subjective methods at all, most teams will perform this testing only at very few critical milestones in a project. They cannot use these methods for frequent, repeated picture quality measurements to diagnose picture quality problems; optimize product design or system performance; and conduct extensive product, system, or content verification.

As a result of these challenges, engineering, maintenance, and quality assurance teams are more frequently turning to instruments that make objective picture quality measurements. There are three major objective picture quality measurement groups: full-reference, reduced-reference, and no-reference.

Full-reference measurements compare a reference video sequence and a test video sequence. In the standard case, the test video is a processed version of the reference video, where the processing has introduced differences between the reference and test videos. No-reference measurements operate only on test video sequences. Reduced-reference measurements base picture quality assessments on extracted properties of the reference and test videos rather than a pixel-by-pixel comparison.

Full-reference objective picture quality measurements most closely correspond to subjective picture quality assessments. When possible, engineers should opt for the more capable full-reference measurements.

NEXT: PSNR Measurements

Page 1: next page

Pages: 1 2 3 4 5 6 7