Corner detection using spline wavelets
Andrew K. Chan, Charles K. Chui, Jun Zha, Q. Liu · Proceedings of SPIE, the International Society for Optical Engineering/Proceedings of SPIE · 1992
Detection of corners in an image is very useful in computer vision and pattern recognition. The existing algorithms for corner detection seem to be insufficient in many situations. The corner detection algorithm proposed in this paper is based on spline-wavelet decompositions. Corner and edge detectors are constructed from the 2-D wavelet transform coefficients. A somewhat sophisticated thresholding technique is applied to remove noise and minor irregularities in the images. Noise can be further reduced if additional processing is applied to the component images at all resolutions. Information on the edges and corners is contained in the component images in all the octaves to facilitate precise localization. A real-time wavelet decomposition algorithm is developed for the corner and edge detectors. It is very efficient and requires very little memory, since most of the computations involve only simple moving average operations and sub-sampling.