Plan-Guided Analysis of Noisy Dynamic Images

Abstract

A wide variety of processes in biology, medicine, or engineering are recorded on video tape or cine film aa moving images. Computer analysis of a moving image, however, requires too much time, if whole areas of a large number of consecutive frames are examined in detail. This paper describes how a plan-guided analysis, which utilizes both a priori and acquired knowledge about the dynamic image, can effiCiently and correctly extract necessary information from the video tape or cine film image. It first analyzes a coarse picture and makes a plan, which is then used in the next fine data collection and processing of the record; the plan specifies (1) frames to be sampled, (2) regions to be analyzed, covering important portions of the dynamic image, and (3) suitable operators and their parameters to be applied to the regions. The effectiveness of the method is shown by an example that analyzes a sequence of noisy radiographic images, a cine-angiogram, which needs to reliably detect curves of specified properties in a large number of frames.