Publications-10

High Resolution Image Processing by SHV.pdf. (1.04MB)
by I. Overington, (25 pages) 2008.

Literature studies during 2008 have led to the conclusion that virtually all conventional image processing endeavouring to define objects in a scene is still carried out such that the first stage in processing is the sensing of individual edge pixels as scalars to the nearest whole pixel. It is evident that there is an increasing interest in ‘vectorization’ of sensed edges from such scalar data, but it seems that resulting vector representations are little more than derivation of best smooth fits to strings of (nearest whole pixel) data points by mathematically defined lines & curves.

By comparison the SHV image processing suite (developed over the past 20 years or more and now available in simplified form as shareware) directly determines the sub-pixel position of edge fragments within individual pixels as sub-pixel vectors. As such, it has been widely demonstrated that greatly enhanced representation of such things as object shape, local & global image motion and local & global relative stereo depth can be achieved.

This report provides a broad discussion of a number of aspects of image processing and serves to provide detailed (page & figure) references to substantial discussion of methods employed & performance capabilities as published in my book ‘Computer Vision: a unified, biologically inspired approach’, Elsevier North Holland, 1992 and other documents.