Use of Second-Order Statistics in Texture Synthesis-by-Analysis

Synthesis-by-analysis is a well established paradigm for reproduction of textures having random appearance. In this contribution, in order to simplify the overall procedure in the case of colored textures, we exploit the second-order correlation existing between the texture chrominance and luminance components. Specifically, we perform the synthesis exclusively on the luminance component, while the chrominance components are reproduced by applying properly designed Wiener filters, whose coefficients have been previously calculated in the analysis stage. This approach yields excellent results especially on natural textures, which intrinsically present a high degree of correlation between the luminance and chrominance components. Moreover, to enhance the synthesis resolution, the same paradigm is then applied to reproduce high-order components of a texture multi-resolution representation based on the Circular Harmonic Functions expansion.

[1]  Anil K. Jain.  Fundamentals of Digital Image Processing , 2018, Control of Color Imaging Systems.

[2]  J. Cadzow,et al.  Image texture synthesis-by-analysis using moving-average models , 1993 .

[3]  Alessandro Neri,et al.  Blind image deconvolution using Bussgang techniques: Applications to deblurring and texture synthesis , 2007 .

[4]  Phil Brodatz,et al.  Textures: A Photographic Album for Artists and Designers , 1966 .

[5]  Jana Reinhard,et al.  Textures A Photographic Album For Artists And Designers , 2016 .

[6]  Gaetano Scarano,et al.  Cumulant series expansion of hybrid nonlinear moments of complex random variables , 1991, IEEE Trans. Signal Process..

[7]  Patrizio Campisi,et al.  Blind image deblurring in the edge domain , 2004, 2004 12th European Signal Processing Conference.

[8]  Patrizio Campisi,et al.  A multiresolution approach for texture synthesis using the circular harmonic functions , 2002, IEEE Trans. Image Process..

[9]  Gaetano Scarano,et al.  Bussgang-zero crossing equalization: an integrated HOS-SOS approach , 2001, IEEE Trans. Signal Process..

[10]  Gaetano Scarano,et al.  Cumulant Series Expansion of Hybrid Nonlinear Moments of /et , 1993, IEEE Trans. Signal Process..

[11]  G. Scarano,et al.  Fractionally spaced Bussgang equalization for correlated input symbols , 2003, 2003 4th IEEE Workshop on Signal Processing Advances in Wireless Communications - SPAWC 2003 (IEEE Cat. No.03EX689).

[12]  B. Julesz,et al.  Human factors and behavioral science: Textons, the fundamental elements in preattentive vision and perception of textures , 1983, The Bell System Technical Journal.

[13]  Alessandro Neri,et al.  Texture synthesis-by-analysis with hard-limited Gaussian processes , 1998, IEEE Trans. Image Process..

[14]  Patrizio Campisi,et al.  Multichannel blind image deconvolution using the Bussgang algorithm: spatial and multiresolution approaches , 2003, IEEE Trans. Image Process..

[15]  Alessandro Neri,et al.  Reduced complexity modeling and reproduction of colored textures , 2000, IEEE Trans. Image Process..