Infrared target tracking with adaptive bandwidth Mean Shift

Tracking objects in image sequence using Mean Shift algorithm has attracted considerable attention. In the paper, an improved Mean Shift tracking algorithm for infrared target of dynamic scale is proposed. In order to reduce the localization errors of infrared targets tracking and suppress the background interference, the proposed method computes the weighted coefficients of the background pixels around the target and represents the model and the candidate in terms of background weighted histogram and color histogram, respectively. After obtain the accurate position of the infrared targets, the paper designs an adaptive bandwidth algorithm for tracking dynamic scale infrared targets. The bandwidth adjusts size along with the change of infrared target scale through make the objective function minimum which is based on similarity of background and target. Experimental results demonstrate that the proposed algorithm is effect and robust to track infrared targets whose size change drastically and rapidly during the sequence, where the literate algorithm obviously fails.