1. Introduction
Medical images obtained from different modalities are widely used in many clinical and research applications. In this paper, a new normalization method using relative contrast to utilize the patient specific information is presented for medical image segmentation problems. For multispectral images, it is widely known that normalization is essential to combine the information provided from different modalities, and schemes such as z-score or energy normalization are classically applied. However, the classical normalization methods normalize different image types by transforming the intensity values into a comparable range, without taking into account the considerable differences between patients. In order to normalize across different subjects, the proposed method uses relative contrast to mimic the manual segmentation procedures that the human readers perform who are essentially comparing the contrast between two classes without being given the actual intensity values.