Correlation Coefficients and Adaptive Threshold-Based Dissolve Detection in High-Quality Videos

Rapid enhancements in Multimedia tools and features day per day have made entertainment amazing and the quality visual effects have attracted every individual to watch these days' videos. The fast-changing scenes, light effects, and undistinguishable blending of diverse frames have created challenges for researchers in detecting gradual transitions. The proposed work concentrates to detect gradual transitions in videos using correlation coefficients obtained using color histograms and an adaptive thresholding mechanism. Other gradual transitions including fade out, fade in, and cuts are eliminated successfully, and dissolves are then detected from the acquired video frames. The characteristics of the normalized correlation coefficient are studied carefully and dissolve are extracted simply with low computational and time complexity. The confusion between fade in/out and dissolves is discriminated against using the adaptive threshold and the absence of spikes is not part of the case of dissolves. The experimental results obtained over 14 videos involving lightning effects and rapid object motions from Indian film songs accurately detected 22 out of 25 gradual transitions while falsely detecting one transition. The performance of the proposed scheme over four benchmark videos of the TRECVID 2001 dataset obtained 91.6, 94.33, and 92.03 values for precision, recall, and F-measure respectively.