关键词: 感知聚类, 张量投票算法, 分形维数, 尺度参数

Abstract: Tensor voting algorithm with strong robustness and non-iteration is a common computing method in perceptual grouping and it is wildly used in image processing. Selecting the scale parameter adaptively in tensor voting algorithm plays an important role in creating the voting fields. In this paper, the scale parameter is selected by fractal dimension. The relationship between scale parameter and fractal dimension is established and then, the adapted tensor voting algorithm based on fractal dimension is proposed. The proposed method is applied in line feature extraction and edge inpainting. The better experimental results are obtained in line feature extraction image and edge repair by the comparison of the proposed method and conventional tensor voting algorithm.

Key words: perceptual grouping, tensor voting algorithm, fractal dimension, scale parameter