Abstarct: Health monitoring and medical diagnoses are impossible without considering different images of organs and tissues. Further, those methods that focus on brain have a great importance because of role of brain in decision-making and control of any conscious and unconscious activity of humankind. Improper functioning of the brain can cause many organic, behavioral and mental problems. Functional magnetic resonance imaging (fMRI) is a new and efficient method for investigation about function of brain through any emotional, behavioral, mental and cognitive task. Analysis and process of fMRI images discriminate between active and inactive regions of brain for any sensor or motor stimulus. Two major kinds of analysis are model-driven (baxsed on a default model) and data-driven (baxsed on structure of input data) methods. In this thesis, with extending the independent component analysis (ICA) method using FastICA and Infomax as two common version of it, a novel idea baxsed on mathematical morphology and Top-Hat transformations has been applied for improving and control the contrast of fMRI images. At first we must do some time-consuming preprocessing steps like: slice-timing, realignment, registration and normalization. Higher degree of performance is achieved testing this method on a huge databaxse of raw fMRI images for an emotional task. Quantitative and qualitative parameters and indices are used to confirm this modification in better recognition of active areas in human brain.