Recognition of similar shapes
Juan M. Simon, Rodolfo M. Echarri · Proceedings of SPIE, the International Society for Optical Engineering/Proceedings of SPIE · 2001
To recognize patterns, correlation and deconvolution techniques are frequently used. The deconvolution yields an identification signal with maximum discrimination and extreme sensibility for small fluctuations in intensity and shape. The phase only filter has the advantage of been more easy to implement in laboratory and it can yield a recognition signal even when object and filter are only slightly similar. The autocorrelation exhibits the latter behavior to a greater extent. In some cases what is of interest is not the equality but the similarity of shapes, e.g., when objects which appear in a fragmented way are to be recognized. In the present article a method to adjust the degree of acceptable similarity between filter and object is proposed and the object is recognized by means of a simple change in a coefficient appearing in the numerical processing.