Gradual transitions are highly confused in high quality videos with camera motion and illumination effects and most of the time are missed or falsely detected. Today’s multimedia technology have immensely advanced that video creators use various eye straightening techniques to attract audience thus making difficult to isolate gradual transitions from such videos. The special effects are confused with actual transitions and a highly robust system would possibly fail to locate the actual endpoints of gradual transitions. We propose a simple and efficient method to detect gradual transitions especially fade in and fade out in videos using mean of features obtained from Multilevel Wavelet transform of the grayscale frames. These values give better correlation between successive frames and the significant decreasing and increasing slopes obtained when properly analysed using global minima and local maxima’s can accurately find the locations of fade out and fade in respectively. An experimentally found threshold able to lower the computational complexity and conserve computation time. We tested our results with manually found transitions extremes points and found that our method worked better than manual approach and possess the ability even the transitions occurred under worst illumination.

Gradual transitions, fade in, fade out, multilevel wavelet transform, global minima, and local maxima