A.L. Priorov, I.V. Apalkov, N.B. Gerasimov, E.A. Pavlov
Assessing the quality of video sequences compressed by Motion JPEG2000 standard is actual due to the low level of correlated with the image distortions, caused by attenuation and quantization of high frequency wavelet coefficients, detection by modern popular quality metrics. JPEG2000 compression algorithm leads to blurring and ringing distortions. The purpose of this paper is to evaluate the quality of video compressed with standard Motion JPEG2000, by measuring these distortions.
Based on the properties of Daubechies 9/7 biorthogonal wavelet basis, the impact of the compression may be considered only on image rows. For distortion detection we need map of edges, obtained by Sobele filter. For each edge in row calculated local blur metric as distance between the left and right extremum of pixel intensity. The blur metric is defined as the average of all local blur metrics. Detecting edges in original or distorted frame makes this metric full-reference (RB) and non-reference (NRB), correspondingly. Local ringing metric is calculated as the sum of multiplications of ringing width on the difference between the maximum and minimum of pixels intensity difference to the left and right of the edge. Ringing metric, computable as the average over all local metrics ping metric is defined as the total ringing metric (TR), and where the local ringing metric is different from 0 actual metric call (AR).
Analyses of proposed video quality metrics follows the methodology recommended by the International Telecommunication Union (ITU-R BT.601). NRB results using nonlinear normalization evaluation practically coincided with the RB. Each of proposed metrics increases with increasing of compression ratio. With compression rate more than 50 appears the effect of global intensity flicker. Proposed NRB has high correlation (95%) with the level of blur defined in test methodology, and proposed ringing metric completes the analysis of distortion in video sequences and makes the quality assessment of video compressed standard Motion JPEG2000 more adequate.