Variable Hough transform for high-resolution line detection

The ability of the Hough transform depends on the quantization of the image space and the Hough space. If the quantization of the Hough space is fine, each voted point on the Hough space may be different, and the line extraction operation may be erroneous because steep peaks are not constructed. Inversely, if the quantization is rough, the accuracy of the obtained parameters also is rough although the extraction itself may work well. This paper discusses the influence of the quantization and proposes a variable Hough transform-a new high resolution Hough transform method. A voted value at the point (ρ, θ) on the Hough space is transformed into a value distribution by the variable filter, and voting operation is performed according to the distribution. The shape and size of the variable filter changes dynamically, with the quantization size and the line parameters (ρ, θ). Statistical experiments show that the variable Hough transform can improve extraction performance and reduce the mean error of extracted line parameters even with noisy images.