Comparison of Video Shot Boundary Detection

Many algorithms have been proposed for detecting video shot boundaries and classifying shot and shot transition types. Here we are using four different methods for comparison, using GIST1, Segmentation, Color Histogram and Motion Activity Descriptor. The Color histogram is the most simplest and fastest method, it computes a histogram and detects a shot based on the comparison of the histogram of each frame but this is sensitive to luminance and motion. The segmentation method uses SVM classifier for the detection of the shot boundaries, these classifiers trained to detect the transition but may be biased to treat all frames as non transition frames which shows its disadvantage. The motion activity descriptor method computes the motion vector for each frame and based on these it detects the shot boundaries but this can be a complex and lengthy processor and may produce incorrect results. The GIST method uses the color as well as gist property of a frame to compute the shot boundaries. The color shows good results for the detection of gradual transitions while the gist is efficient to detect the abrupt cuts. Thus we can say the GIST is the best method after carrying out comparison between all these methods.
1. Gist means Gesture Interpretation Using Spatio-Temporal Analysis, where Spatio-Temporal means which has both space as well as time properties like the movement of hand which shows the variation in both space as well as time.