The optical model of human joints synovial fluid is proposed. The statistic (statistic moments), correlation (autocorrelation function) and self-similar (Log-Log dependencies of power spectrum) structure of polarization two-dimensional distributions (polarization maps) of synovial fluid has been analyzed. It has been shown that differentiation of polarization maps of joint synovial fluid with different physiological state samples is expected of scale-discriminative analysis. To mark out of small-scale domain structure of synovial fluid polarization maps, the wavelet analysis has been used. The set of parameters, which characterize statistic, correlation and self-similar structure of wavelet coefficients’ distributions of different scales of polarization domains for diagnostics and differentiation of polycrystalline network transformation connected with the pathological processes, has been determined.