Likelihood constrained smoothing interpolation using segmentation information

Interpolation techniques are required in many areas of medical image analysis. Due to the discrete nature of all medical images, interpolation is a necessary step in resampling, which is required for image registration and all geometric image manipulations. In this paper, a novel smoothing interpolation technique is proposed. This method incorporates segmentation information using statistical approach. The interpolation is constrained by the likelihood of each data point belonging to a certain segmentation class. Testing was done on the brain MRI images, and very promising results were shown. The ultimate goal of this interpolation technique is to apply it in medical image registration for increasing its accuracy