Basic Contents

Computer Vision ... (Published 1992)

Simulated Human Vision..... Ian Overington

Location: Eastbourne. UK
ianoverington@simulatedvision.co.uk ............ www.simulatedvision.co.uk

8.2. Inter-pixel noise.
8.2.1. 2-D discrete sampling noise.
8.2.2. 1-D discrete sampling noise.
8.2.3. Discrete chromatic noise.
8.3. Spatial interpolation.
8.4. Temporal interpolation.
8.5. Conclusions.

Chapter 9. General Flow Field Analysis.


In Chapter 5 the possible methods of sensing local fragments of motion were discussed. In
particular, a simple method was described, whereby sub-pixel estimates of the components of motion
orthogonal to local edge details could be derived directly (that is, without the need for feature
matching), on a pixel to pixel basis, from
pairs of frames in a temporal sequence. Taken together
with parallel measurements of the local
orientation of edges, on a pixel to pixel basis, obtained from
mean of the two frames, essentially as described in Chapter 4, one therefore had available a
vernier map of orthogonal components of motion vectors over an entire image. Given such a map,
it is the purpose of this Chapter to show how such fragmentary data may be used to generate, simply
and absolutely, a variety of attributes of the underlying flow field. These include the motion of local
bodies relative to a fixed background, global motion due to lateral movement of the imaging sensor,
local motion within global motion, and the characteristic differential and progressive motion due to
forward sensor motion. This latter will be shown to lead to a simple and ready method for
determination of the focus of expansion (FOE) (i.e. the point to which the sensor platform is