Prostate segmentation from MRI image is a challenging task due to the appearance variations of different patients and in particular feature and density heterogeneity within a same image. To address these challenges, we propose a collaborative learning based model to adaptively choose the optimal features from a set of feature candidates to improve prostate boundary delineation. In our method, on the basis of weighted multi-view collaborative clustering, our proposed new ranking and selection scheme automatically determines the unique feature for optimally contouring individual image regional sectors that exhibit density inhomogeneity and feature variations. The approach was tested on prostate MR scans without endorectal coil from PROMISE12. The segmentation accuracy was evaluated with respect to five most often used evaluation criteria in terms of spatial volume overlap and shape similarity. The experimental results demonstrated that our proposed method based on collaborative learning model outperformed graph-based algorithms and the multi-view clustering models.