Performance measurement for a wavelet transform-based video compression

  • Authors:

  • Abinashi Dhungel;Michael Weeks

  • Affiliations:

  • Georgia State University, Atlanta, Georgia;Georgia State University, Atlanta, Georgia

  • Venue:
  • Proceedings of the 49th Annual Southeast Regional Conference
  • Year:
  • 2011

Quantified Score

Hi-index 0.00

Visualization

Abstract

A wavelet transform-based video compression algorithm consists of i) 3D wavelet transform ii) Quantization, and iii) Coding. Since wavelet analysis has similarities with characteristics of human visual system, it is desirable to use a quality metric that agrees with human perceptual evaluation. One of the widely used quality metric for the reconstructed image and video is the peak signal to noise ratio (PSNR). The disadvantage of PSNR is that the result does not correlate very well with subjective or human evaluation of the perceived quality. The Structural Similarity (SSIM) is a quality metric that quantifies quality based on the changes on the structural information variation of the image due to error signal rather than the visibility of the error signal and performs better than PSNR in terms of perceptual correlation with human observer [17]. This paper describes an implementation of a 3D wavelet transform-based video compression model and a new video quality metric which is an extension of the SSIM to the third dimension. Experiments show that the new metric is perceptually more accurate than the PSNR for measuring video quality.