next up previous
Next: Hough Transforms Up: Object Recognition Previous: Invariants

Pattern Recognition

One way to spot if an object (or part of an object) is to look for a pattern of primitives occurring in a scene.

Template Matching

Template matching is a simple in which instances of prestored patterns are sought in an image.

Template matching has been performed at the pixel level and also on higher level.

Pixel Level Template Matching

Here we seek low level pixel templates. There are 4 approaches:

Total templates
-- Here an exact match is sought. The Template is the same size as the input image. There is no rotation or translation invariance.
Partial templates
-- Here the template is free from the background. Multiple matches are allowed. Partial matches may also be allowed. Care must be taken in this case -- an F template could easily match to an E. Storage requirements may limit pattern representations.
Piece templates
-- Here patterns to matched are broken into component templates. E.g. The pattern A could be recognized by 3 templates /, tex2html_wrap_inline8631, - ANDed together. Order in which component templates matched is important -- largest first. Storage requirements less in this method.
Flexible templates
-- These templates can handle stretching, misorientation and other possible deviations. A good prototype of a known object is first obtained and represented parametrically. As other examples are presented then the parameter are modified (parameter adjusted learning). Relaxation based methods used to determine a possible match.

High Level Template Matching

A problem with pixel based is that although fairly cheap and simple to implement to rotation and translation is a problem. Also images are rarely perfect suffering form blurring, stretched and other distortions and peppered with noise.

E.g. How could low level methods cope with handwritten characters?

High level template matching methods operate on an image that has typically been segmented into regions of interest.

Regions can be described in terms of area, average intensity, rate of change of intensity, curvature and also compared -- bigger than, adjacent to, above, distance between.

Templates are described in relationships between regions. Production rules and other linguistic representations have been used here. Also statistical methods (relaxation based techniques) have been applied to perform the matching.


next up previous
Next: Hough Transforms Up: Object Recognition Previous: Invariants

dave@cs.cf.ac.uk