Stephens, Greg ; Mora, Thierry ; Tkacik, GasperIST Austria ; Bialek, William
The scale invariance of natural images suggests an analogy to the statistical mechanics of physical systems at a critical point. Here we examine the distribution of pixels in small image patches and show how to construct the corresponding thermodynamics. We find evidence for criticality in a diverging specific heat, which corresponds to large fluctuations in how "surprising" we find individual images, and in the quantitative form of the entropy vs energy. We identify special image configurations as local energy minima and show that average patches within each basin are interpretable as lines and edges in all orientations.
Physical Review Letters
This work was supported in part by NSF Grants No. IIS-0613435, No. IBN-0344678, and No. PHY-0957573, by NIH Grant No. T32 MH065214, by the Human Frontier Science Program, and by the Swartz Foundation. CC BY 3.0
Stephens G, Mora T, Tkacik G, Bialek W. Statistical thermodynamics of natural images. Physical Review Letters. 2013;110(1):018701. doi:10.1103/PhysRevLett.110.018701
Stephens, G., Mora, T., Tkacik, G., & Bialek, W. (2013). Statistical thermodynamics of natural images. Physical Review Letters, 110(1), 018701. https://doi.org/10.1103/PhysRevLett.110.018701
Stephens, Greg, Thierry Mora, Gasper Tkacik, and William Bialek. “Statistical Thermodynamics of Natural Images.” Physical Review Letters 110, no. 1 (2013): 018701. https://doi.org/10.1103/PhysRevLett.110.018701.
G. Stephens, T. Mora, G. Tkacik, and W. Bialek, “Statistical thermodynamics of natural images,” Physical Review Letters, vol. 110, no. 1, p. 018701, 2013.
Stephens G, Mora T, Tkacik G, Bialek W. 2013. Statistical thermodynamics of natural images. Physical Review Letters. 110(1), 018701.
Stephens, Greg, et al. “Statistical Thermodynamics of Natural Images.” Physical Review Letters, vol. 110, no. 1, American Physical Society, 2013, p. 018701, doi:10.1103/PhysRevLett.110.018701.
IST-2016-401-v1+1_1281.full.pdf 416.96 KB
Link(s) to Main File(s)