Image Quality Considerations of Compressed Video Imagery

Abstract

Most work to date on image data compression has been based on single static images and the use of mean square error (MSE) as an objective quality criterion. However, many current requirements for video bandwidth compression, such as Remotely Piloted Vehicle and "smart" ordnance video links, include dynamic (moving) scenery and severe channel noise. The image degradation effects of the bandwidth compression technique will then fall into three categories: Degradation of a static image; effects due to any frame segmenting and partial frame transmission used; and the impact of uncorrected channel errors. In this environment, the more powerful, but also more complex techniques, i.e., the various transform techniques, may provide less acceptable imagery than simple techniques (such as DPCM or Lockheed's CAQ algorithm) since their complexity may require frame slicing, i.e., the compression and transmission of only a segment of a frame during each frame period. Also, the one-dimensional techniques often prove more effective than twodimensional techniques because the vertical redundancy they retain can be used to eliminate uncorrected errors. The MSE criterion for image quality, known to be crude and often misleading in static images, is totally inadequate to quantify the impact of scene dynamics and uncorrected channel errors.