Performance Analysis of Video Magnification Methods

Abstract

Video is a huge source of information and the only way to extract that information is video processing. Depending upon the information to be retrieved or analyzed, the video processing techniques are applied. Some of the information consists of imperceptibly small motion, color changes, or sound changes. To estimate the small motions, which are not visible to naked eyes, various video magnification techniques are used. In this paper, various video magnification techniques- Eulerian video magnification, Phase-based video magnification with both Laplacian and Gaussian pyramid are implemented and their comparative performance analysis is presented. The various performance metrics used are PSNR, amplification factor, execution time, intensity of color and motion magnification. The effect of changing the level of pyramid and amplification factor on the performance of particular method is also analyzed. With increase in the level of pyramid quality of magnification reduced, while for higher amplification factor (>200) output exploited drastically. Results reveal that the Eulerian video magnification method is fast and simple but supports small amplification factors while the phase-based method is very slow and complex but supports the large amplification factors. To get better performance of the video magnification method, it is required to develop a new method using hybrid approach and its scope can be further extended in biomedical field to measure the biological parameters.