helloTC / texture-synthesis-portilla-simoncelli

Texture Sysnthesis by Portila and Simoncelli

Python3 implementation of Polltila and Simoncelli(2000)'s texture synthesis.

Parametric Texture Model Based on Joint Statistics of Complex Wavelet Coefficient, Portilla, J. and Simoncelli, E.(2000)

Basically this repository is a port of Matlab programs of NYU. But "steerable pyramid" part is based on Briand et al.(2014) .

The Heeger-Bergen Pyramid-Based Texture Synthesis Algorithm, Briand,T. et al. (2014)

I have already published the python3 implementation of steerable pyramid by their equations.

Results

Attentions !!
All images(except results) are from NYU website. The copyrights belongs to it.

In this implementation, the "real version" of steerable pyramid is used. But the results are very close to the ones of "complex version" of Portilla and Simoncelli(2000). See bellow.

Gray Scale

Original image

Synthesized image

Original image

Synthesized image

Color Version

Original image

Synthesized image

Original image

Synthesized image

Usage

Environment

• python3.5 (3.0+)
• GPU is not used.

Execution(Gray Scale Version)

• create 'out' directory.
• python texture_synthesis_g.py -i radish-mono.jpg -o out -n 5 -k 4 -m 7 --iter 100, if you want to know details about parameters, see source code

Execution(RGB Color Version)

• create 'out' directory.
• python texture_synthesis_g.py -i radish.jpg -o out -n 5 -k 4 -m 7 --iter 100, if you want to know details about parameters, see source code

Acknowledgement

thanks to the authors of NYU.

• Matlab programs of NYU

References

• Parametric Texture Model Based on Joint Statistics of Complex Wavelet Coefficient, Portilla, J. and Simoncelli, E.(2000)
• The Heeger-Bergen Pyramid-Based Texture Synthesis Algorithm, Briend, T. et al.(2014)