Shot Boundary Detection based on Multilevel Difference of Colour Histograms

Shot boundary detection (SBD) is a fundamental step in video retrieval and analysis. Although the existing methods have achieved a lot on this field, this problem has not been sufficiently resolved, especially for the gradual boundary. This paper proposed a new method for SBD using a three-stage approach based on the Multilevel Difference of Colour Histograms. In the first stage, we detect the cut candidate boundaries and gradual candidate boundaries using two self-adapted thresholds. In this stage, we transform the candidate gradual change to cut change so that the cut change and gradual change can be dealt in a same way. It helps to keep a high recall rate. The second stage detects the local maximum difference of the MDCH generated by shot boundaries to filter the noises incurred by object motion, flash and camera zoom. A voting mechanism is used to make a final detection in the third stage. The second and the third stage contribute to a high precision rate. We evaluated the proposed method on TRECVID datasets and compared it with other methods. Experiment results demonstrate that the proposed method achieves high detection accuracy with low computational cost.