Depth Dependent Correction of Feature Point Extraction Algorithms for Optical Measurement Systems

Tobias Hanning


feature extraction, point spread function, depth depended distortion correction


In computer vision the blurring of observed objects in an image is modeled by a convolution with a point spread function (PSF). The extraction of basic image features like points, lines, circles or ellipses is done generally without knowing the PSF exactly. Therefore, point extraction must result in an erroneous positions. In this article we show that even for first order optics the point spread function is not symmetrical. Furthermore, it depends on the position of the observed point. Therefore, for optical measurement systems any feature point extraction should be corrected with respect to this position. In this article we show how this correction can be coded and calibrated within a standard camera model. Several experimental results show the gain of the proposed correction.

Important Links:

Go Back