
Title: Statistical Thermodynamics of Natural Images Authors: Greg J. Stephens, Thierry Mora, Gasper Tkacik, and William Bialek Ref: PRL 110, 018701 (2013) http://prl.aps.org/abstract/PRL/v110/i1/e018701 To understand how the brain efficiently processes visual information, it is important to determine the statistical properties of natural images. Interestingly, natural images exhibit scale invariance, reminiscent of phase transitions encountered in statistical physics. In this paper, the authors make the connection explicit, and find evidence for criticality in the probability distribution of natural images. For instance, the diverging specific heat in thermodynamics here corresponds to large fluctuations in our "surprise" to any given scene, which makes the compression of large natural images a difficult task.

Title: Statistical Thermodynamics of Natural Images
Authors: Greg J. Stephens, Thierry Mora, Gasper Tkacik, and William Bialek
Ref: PRL 110, 018701 (2013)
http://prl.aps.org/abstract/PRL/v110/i1/e018701

To understand how the brain efficiently processes visual information, it is important to determine the statistical properties of natural images. Interestingly, natural images exhibit scale invariance, reminiscent of phase transitions encountered in statistical physics. In this paper, the authors make the connection explicit, and find evidence for criticality in the probability distribution of natural images. For instance, the diverging specific heat in thermodynamics here corresponds to large fluctuations in our "surprise" to any given scene, which makes the compression of large natural images a difficult task.