The improved progressive probabilistic hough transform for paper wrinkle detection

Abstract

Wrinkles are the typical paper defects that usually appear in paper production. Because wrinkle has line patterns in paper image, it can be detected through Hough Transform (HT) based methods. Among those methods, Standard Hough Transform (SHT) is a robust line detection approach, but the long time computation handicaps its application in real-time detection. Progressive Probabilistic Hough Transform (PPHT) is the improved version of SHT, however, when it is used for the images with heavy noise, the accuracyreal-time detection system rate of detection is not high. In order to apply HT-based methods to real-time detection system and accomplish high detection performance as well, this paper proposes two improvements in PPHT, including segment-weighted voting and density-based segment filtering. The experiments on real paper images with wrinkles demonstrate that the improved PPHT is valid and the detection performance is much better than the traditional PPHT.