A local soft voting method for texture-based vanishing point detection from unstructured road images

Vanishing point estimation algorithms based on Gabor filters have been widely published. By applying Gabor filters, the texture orientation and the confidence level at every pixel of the image can be estimated. An adaptive soft voting method to obtain a vanishing point was proposed in other research. Although that method can detect a proper vanishing point in most cases, its computational cost is very high. The goal of this paper is to reduce the computational cost of the algorithm. For this purpose we herein propose a new local soft voting method by scanning voters having high confidence levels.