This research work introduces a simple yet effective method for brain tumor detection using proposed dataset of 1500 images. There are different types of brain tumor; among the existing we have considered four different types i.e. CNS Lymphoma, Glioblastoma, Meningioma, and Metastases. The four major steps in the proposed method are pre-processing, segmentation, post-processing and image fusion. In the pre-processing, 2D-Adptive filter is applied to enhance the quality of the image. Otsu's segmentation is used to extract tumor region from normal tissues. The segmented region contains skull boundaries in the form of noise; hence morphological operations i.e. erosion and dilation have been applied to remove the extra noise caused by segmentation. Overlay based image fusion is applied to get a clear visual of segmented tumor region. We achieved a detection rate of 93 percent with 7 percent error rate using this dataset. Furthermore, we classify the tumor into benign and malignant based on the size of tumor.