High speed line detection by Hough transform in local area

In line detection method first let the candidates of line segment whose positional precision or reliability are guaranteed to some extent be extracted in a local area, then let the line be extracted by a verification process in the existing area of candidate line which is defined by an extension process for the line segment in the pattern space, and finally let all lines be extracted by a shifting operation of the local area in the image appropriately. By these processes, our method realizes the absorption of the quantization errors in Hough transform, as well as the extraction of short line segments. Furthermore, parallel processing is available for the hardware implementation. In this paper, we show theoretically and experimentally that calculation costs for line detection is extremely reduced without any degradation of detectability and reliability in comparison with a conventional Hough transform.