A novel gradient based algorithm to correct the intensity inhomogeneity of MR images

作者：Li Zhaohui; Ling Qiang; Yu Yi

来源：Journal of Computational Information Systems, 2015, 11(6): 2155-2162.

DOI：10.12733/jcis13797

• Bias field
• Bias-correction methods
• Gradient based algorithm
• Intensity inhomogeneity
• Intensity inhomogeneity correction
• Low-order polynomials
• Polynomial fittings
• Reconstructed image

This paper proposes a novel intensity inhomogeneity correction algorithm for magnetic resonance images. The intensity inhomogeneity is modeled as a multiplicative and slow-varying surface. Using a specified gradient operator based on Gaussian kernel, the algorithm first extracts the slowly varying gradient from the original image, then reconstructs the signals by discarding small gradient values in the image caused by the bias field. The algorithm then solves for the intensity inhomogeneity by fitting a low-order polynomial function to minimize the square error between true image and the reconstructed image. The algorithm was tested on both simulated and clinical magnetic resonance images and is shown to well estimate the bias field and restore the homogeneity of MR images of different body parts. The results indicated that the algorithm can achieve better or comparable performance as some of the most commonly used bias correction methods.

• 出版日期2015
• 单位中国科学技术大学; 中国科学院