Intensity Non-uniformity Correction for Image Segmentation

Abstract

Intensity non-uniformity or intensity inhomogeneity usually occurs in Real world Images, those images cannot be segmented by using image segmentation. The most commonly used algorithms in image segmentation are region based and depends on the homogeneity of the image intensities which usually fails to produce accurate segmentation results due to the intensity non-uniformity. In this paper we proposed a novel region based method for image segmentation which can be able to discuss with intensity non-uniformities in image segmentation. First according to the image models with intensity non-uniformities we define a local clustering criterion function for the intensities in the image neighbourhood of each part. The local clustering criterion function is then integrated with respect to the neighbourhood center to give a global criterion of image segmentation. In a level set formulation this criterion defines an energy in terms of level set functions that represents the partition of image domain and a bias field that corresponds to the intensity non-uniformity of the image. Therefore, by minimizing the energy we can able to segment the image simultaneously and estimate the bias field can be used for the intensity non-uniformity correction. This method is applied on MRI images and real world images of various modalities with desirable performance in the presence of intensity non- uniformities. The experiment results show that the method is stronger, faster and more accurate than the well- known piecewise smooth model and gives promising results. As an application this method is used for segmentation and bias correction of real world images and MRI images with better results. Keywords -Bias field, Energy minimization, Image segmentation, Intensity non-uniformity, Level set method.