The automatic detection and extraction of tumor area in Magnetic Resonance Imaging (MRI) is an important and challenging task. This paper presents a fully automatic and unsupervised method for fast and accurate extraction of brain tumor area from MR images. The proposed method named as Saliency Based Segmentation (SBS) is based on visual saliency. The saliency model detects the pathologically important area and then fuzzy thresholding is used for extraction of the detected region. The performance of SBS is compared with Adaptively Regularized Kernel-Based Fuzzy C-Means Clustering, Mean Shift and Fuzzy C-Means clustering with Level Set Method. The experimental evaluation validated on BRATS database using Jaccard index (0.84 ± 0.04), Dice Index (0.91 ± 0.02), Execution time (2.99 ± 0.29), Precision (0.82 ± 0.16), Recall (0.97 ± 0.03) and F-measure (0.88 ± 0.10) demonstrates that SBS achieves better segmentation results even in the presence of noise and uneven illumination in images.