© 2018 IEEE.Brain cancer is one of the most fateful diseases today. Early diagnosis is of great importance in the treatment of this disease. To accomplish a fast and accurate diagnosis, numerous studies have been performed around the world. In this study, a computer aided tumor detection task is proposed for brain MR images. To prevent over-segmentation a set of methods such as bilateral, gauss, order statistics filters, morphological and sharpening operations are applied for denoising, emphasizing fine details and enhancement steps prior to watershed segmentation. Finally, a rule based elimination is proposed to reduce the false positive detections and increase the performance. Experimental results demonstrate that the proposed method is satisfying to detect brain tumors.