Optic Disk is one of the prominent features in human fundus images. Automatic localization and segmentation of optic disk can help in early diagnosis of diabetic retinopathies and preventing vision loss. In this paper robust method for optic disk detection and extraction of optic disk boundary is proposed based on morphological operations, smoothing filters and markers controlled watershed transform. This method has shown significant improvements in terms of detection and boundaries extraction of optic disk. This method used two types of markers: internal marker and external marker. These markers first modified the gradient magnitude image and then watershed transformation is applied on this modified gradient magnitude image for boundary extraction. The proposed method has optic disk detection success rate of 100% for Shifa and 87.6% for DIARETDB1 databases. Proposed method achieved average overlap of 51.19% for DIARETDB1 database and 73.98% for Shifa database which is higher than currents methods. Experimental results clearly demonstrate an efficient performance of the proposed algorithm.