关键词: 图像处理, 灰度目标跟踪, Gabor-LBP模型, Mean Shift

Abstract: In gray image sequence, due to the sensitivity to illumination variation and the lack of the information for target representation, target tracking is very difficult. This paper proposed a novel tracking algorithm, which integrated the target's Gabor wavelet transform features and rotation invariance uniform Local Binary Pattern (LBP) texture description operator to construct the target's multi-level texture feature models, and used Mean-Shift to track. The algorithm first adopted Gabor wavelet transform to extract multi-scale and multi-orientation features of target to extend the range of feature extraction, and then the rotation invariance uniform LBP operator was applied to encode these features to enhance the validity of the extracted features. Finally, the target's multi-level Gabor-LBP texture feature models were constructed by texture pattern joint probability histograms, and Mean-Shift was adopted to track. The experimental results show that this algorithm can effectively cope with illumination variation, clutter and rotation in gray target tracking.

Key words: Image processing, Gray target tracking, Gabor-LBP model, Mean Shift