Video Quality System

May 29, 2007 -- Researchers at TRLabs have developed a prototype no-reference video quality system to locate and measure the impact of three types of common impairments that influence the quality of television and video signals.

Benefits…

• Locates and measures the impact of the three most common errors in analog / digital signals
• No-reference approach allows for real time video quality analysis
• Average performance of more than 30 fps
Background

The purpose of image quality metrics (IQM) is to accurately assign a photograph or video a numerical value (0 to 5) to reflect its quality. Quality can be thought of as the degree of similarity between the image and its expected form. For instance, a person would consider a blurred image of a house as being low quality because it deviates from the expected appearance of a building. Unable to currently duplicate this sophisticated behavior, a machine must instead base its quality assessment on the statistical and geometrical properties of an image.
The majority of IQMs are full-reference (FR), meaning that in addition to the given or coded image, an “original” or “perfect” version of that photo is assumed to exist. The quality score is then based on extent that the coded image deviates from the perfect version. Alternatively, a no-reference (NR) does not assume that a perfect version of the photo exists, which leads to complexity in the design of the metric.

Research in the area of video quality likely began as an extension of image quality analysis. Techniques to estimate the degree of blocking [2, 5, 6] and blur [2, 3, 4] in still images are steps in this direction. However, these blocking metrics do not take advantage of the temporal aspect of video and several of them involve the Fourier transform and other complex statistical measures, making them unsuitable for real-time video monitoring. Though inventive, the blur metrics are not particularly useful in the context of video analysis since this sort of impairment rarely occurs. Researchers at TRLabs have developed three NR video quality metrics to locate and measure the impact of the three common impairments of television and video signals. For analog signals, the error of interest is the distortion introduced by random noise in the signal. In the case of digital signals, two fundamental types of errors are of interest. The first is the blocking artifact that is so pervasive among DCT-based compression schemes such as MPEG. The second category includes errors caused by random changes to the bit stream of a signal.

Login to see more detailed information on this ETB

See more technology bulletins »