A New Model of Nature Images Based on Generalized Gaussian Distribution

We propose a new statistical model of nature images named 2D joint differential image histogram (JDIH). To simulate this model, we define a kind of 2D generalized Gaussian distribution (GGD) symmetrically in every direction by extending the ID GGD function. The 2D JDIH and its estimated parameters can efficiently measure the images inner and inter correlations of local areas in a special direction or between different directions. Because the correlations in nature images behave like image textures, JDIH and DIH can measure the texture complexity of nature images and are useful in many fields of image analysis.