Image fusion technique using fuzzy sets

摘要

Image fusion techniques are being developed in order to generate an image which is superior to all the input images obtained by using various methods, from certainty and/or resolution points of view. These techniques have numerous applications in image processing, in general, and in biomedical imaging, in particular, where, for the same object various imaging systems produce images that have different characteristics. Fuzzy set theory can be fruitfully used for image fusion since the relation itself between tissues and pixel values is fuzzy by nature and the uncertainty associated with each imaging system can be handled easily; at the same time, the theory provides a large variety of operators for combining the available fuzzy information and is well adapted to image processing [1]. In this paper a new method for multi-modality image fusion is proposed, based on the fuzzy set theory. The relation between tissues and pixel values as well as the uncertainty resulted from noise and systematic measurement errors are defined by using appropriate fuzzy sets. The method presented is based on the computation of the possibility measures of one of these fuzzy sets with respect to the other for each tissue and each image uncertainty. The corresponding values are then processed to produce the final image instead of dealing from the beginning with similar fuzzy membership functions for all the imaging systems. A simulation is also presented to show the suitability and the performance of the method.