The trapezium as an image processing primitive
Neill Campbell, BT Thomas · Bristol Research (University of Bristol) · 1991
Introduces the trapezium as an image processing primitive. A trapezium may represent many individual pixels whilst still being powerful enough to model scenes accurately using comparatively few elements. Most solutions to vision processing problems need to be fast, robust and accurate. However, due to the large amount of data in a typical image, reliable real-time analysis is not always possible. Therefore it is desirable to have a method which will: reduce the quantity of data present, preserve all information relevant to the particular task. The new idea introduced here is that an image can be approximated by a number of horizontal strips, each consisting of several abutting trapezia. The strips are of a fixed-height so as to reduce the complexity of obtaining these trapezia. In this approximate representation, the accuracy can be matched to the application.