April 2012
·
54 Reads
·
2 Citations
This paper describes a new approach to discontinuity detection involving the processing by support vector machine (SVM) of Haar wavelet transforms of local image windows. The method labels each pixel in an image as a pixel of discontinuity or otherwise by processing the Haar wavelet transform of the image in a window about the pixel using an SVM. The pixel labeler is trained using software for labeling training images by a human operator. As a pixel classifier, to within a moderate morphological tolerance, the discontinuity detector has an accuracy of 99% on the images for which it has been tested.