Content-based retrieval of line-drawing images using angular information of hough transform
Pasi Fränti, Alexey Mednonogov, Heikki A. Kalviainen · 2000
We introduce a novel method for content-based retrieval of linedrawing images. The method is based on Hough transform, which is capable of extracting global linear features in the image. Images are indexed by generating feature vectors from the quantized accumulator matrices from the Hough transform. The matrix is first thresholded in order to preserve only the most significant values. The feature vector is then constructed by summing up the significant coefficients in the columns of each θ-value. Thus, only the angular information is used. The approach allows simple implementation of scale, translation and rotation invariant matching.