MR image bias field harmonic approximation with histogram statistical analysis

This study investigates a method to correct the intensity inhomogeneity field of magnetic resonance image. The algorithm takes full advantage of the properties of the magnetic resonance image, namely, the piecewise character (piecewise constant and piecewise smooth) of true image which characterizes a physical property of the tissue anatomical structure and the smoothly varying property of bias field which accounts for the intensity inhomogeneity. An energy function was constructed by embedding this characters into the image model. We can get the estimation of bias field and the segmentation of tissue by minimizing the energy function. The initial parameter of energy function is calculated automatically by statistical analysis. By mixing the fitting basis function with cosine function and polynomial function, we can obtain an accurate approximation of the bias field. A comparative performance evaluation is carried out over a large set of experiments using synthetic magnetic resonance data. Besides, a set of tests on real prostate magnetic resonance image with severe intensity inhomogeneity field is shown to demonstrate the validity of our method.