Shot boundary detection with mutual information

摘要: We present a novel approach for shot boundary detection that uses mutual information (MI) and affine image registration. The MI measures the statistical difference between consecutive frames, while applied registration compensates camera panning zooming. Results different sequences are presented to illustrate motion zoom compensation robustness of illumination changes. Furthermore we show has no effect at boundaries themselves therefore doesn't corrupt result. Because algorithm analyses frames sequentially, its parallelization is straightforward even distributed memory architectures. quantify speed-up on LINUX cluster communicational load implementation almost negligible, resulting in an linear speed-up.