High performance video places severe demands on playback system and display device resources. Motion playback errors such as irregular motion playback and image breakup are common, making accurate measurement of motion critically important for device selection and system tuning to ensure the desired viewing experience. Direct observation of the actual displayed image (screen capture) is the only way to perform a complete end-to-end system test that accounts for all possible sources of motion errors, including original video capture, compression/decompression, playback system performance, and display device operation. Unfortunately for the development of measurement techniques, the fundamentally different ways that different display technologies present imagery can easily confound sensitive measurement techniques, producing measured playback performance differences across multiple display technologies such as LCD (liquid crystal display), plasma, CRT (cathode ray tube), and DMD (digital micromirror device) that are disproportionately large compared to the actual differences (if any) seen by a human viewer. Crosstechnology measurement tools are necessary to ensure the validity of measures across multiple technologies. The methods being used include a combination of test materials (both selected live video clips and synthetic clips), capture technique (including control of capture rate, triggering, and timestamping), and analysis. This paper describes these methods, then gives several illustrative examples of the use of these methods by the project. The NIST Motion Image Quality Measurement project uses a combination of quantitative measures and subjective analysis to evaluate motion imagery and to identify the factors that determine overall performance. The NIST project has conducted tests to measure (1) perceived image quality and (2) motion image interpretability, as a function of factors including compression, video motion and other scene content, playback system performance, and frame rate. Tests have been conducted using a variety of playback systems with different levels of performance, and using a variety of display technologies.
BACKGROUNDNIST is developing new ways to measure the motion characteristics of video playback, as seen by the human viewer. The ability to perform such measurement is important for designing and configuring systems, and authoring content. By definition, motion image quality relates to what a human viewer can see (a subjective response). It is important, however, to be able to produce repeatable numeric measurements (an objective response), to compensate for variations among human viewers and to allow accurate modeling of system performance, for example when tuning a system to meet a given set of requirements. The NIST team has found instances in which the different characteristics of different display technologies, such as LCD, plasma, CRT, or DMD, produce results that look very similar to the human eye, but that are rated very differently from one another using a specific screen ca...