ut ut

Laboratory for Image & Video Engineering

Bivariate Natural Scene Statistics

Introduction

Natural scene statistics (NSS) are important factors both towards understanding the evolution of the human vision system and for designing image processing algorithms. Univariate NSS have proven to be powerful tools driving a variety of computer vision and image/video processing applications. Multivariate NSS models, however, received relatively little attention.

We developed a closed form bivariate correlation model of bandpass and normalized image samples that completes an existing two-dimensional joint generalized gaussian distribution model of adjacent bandpass pixels.

The optimal parameters yielding the best average correlation fit over various spatial orientations in our model can be found here .

The Mean Squared Error (MSE), computed in the case of the amplitude function in our model can be found here , for the peak here , and for the correlation function here .

The χ2, computed in the case of the amplitude function in our model can be found here , for the peak here , and for the correlation function here .

Relevant publication:

  • Z. Sinno, C. Caramanis, and A. C. Bovik, "Towards a Closed Form Second Order Natural Scene Statistics Model," IEEE Transactions on Image Processing , submitted.

Back to Research page