Object Tracking Using Mean Shift Algorithm with Illumination Invariance
Mirza Zeeshan Baig, Aniket V. Gokhale · 2015
Visual tracking is a very challenging task in real life applications because of the instability in illuminations. An illumination or brightness variation is the significant issue in the tracking of object of interest. A satisfying method is proposed in this paper to correct the non uniformity in illumination. In this method discrete cosine transformation (DCT) is used to normalize the variations in illumination in logarithmic domain. High frequency coefficients of discrete cosine transformation are discarded because illuminations are reflected in low frequency coefficients only. We compensate the illumination variations by modifying DC coefficients and taking average of these DC coefficients of neighboring frames of current frame. The corrected illumination videos are applied to mean shift algorithm to track the object of interest.