Non iterated mean shift and particle filtering

In this paper, non iterated mean shift (NIMS) algorithm which uses mean shift algorithm without any iteration is addressed for single object tracking to increase speed of MS algorithm. The proposed NIMS have been combined with particle filter and then speed and accuracy comparisons have been performed with the existing MS-PF method. An isotropic kernel has been used for histograms. The Bhattacharyya criterion has been adopted for similarity measuring of color modeling. Background effects on reference histogram are reduced by corrected-background-weighted histogram (CBWH). Some code optimization suggestions such as MAC processing instead of array processing and zero detection are applied for enhancing the process speed. Results of numerous video samples have demonstrated that this tracking can achieve acceptable speeds.