Multiscale corner detection of contour images using wavelet transform

摘要: A new multiscale corner detection method is proposed based on dyadic wavelet transform (WT) of the orientation function a contour image. As decomposition WT complete and its scales are sparse, all defined as natural for detection. The points that modulus maxima (WTMM) at different taken candidates. For each candidate, sum corresponding normalized WTMM used significance measure "cornerness". utilization information makes performance detector independent to type input images. restricted by length, which algorithm adaptable both long contours short contours. Both subjective objective evaluation illustrate better compared conventional methods.