ROTATION HOUGH TRANSFORM

This article discusses how image space discretization errors affect the standard Hough transform (SHT). It is observed that the discretization errors are different on various columns, that is, they are related to the. values. Furthermore, the columns corresponding to theta=0 degrees and theta=90 degrees are seen to have the smallest errors. Based on this observation, a new Hough transform (rotation Hough transform, RoHT) is proposed. The given image is rotated by.. and the SHT is employed, generating the HT data on the columns corresponding to theta=0 degrees and theta=90 degrees only. After the image is rotated 90 degrees/Delta theta times, the HT columns generated are registered to the HT matrix of the original image. In this way, only theta=0 degrees and theta=90 degrees are used to calculate the votes for each cell in HT space, and the errors are minimized. The experiments verified the performance of the proposed method.