Segmentation of knee joint is an instrumental task in a medical fraternity during knee osteoarthritis (KOA) progression. In general, it is a primary concern to segment, detect and extract the defects from knee magnetic resonance images (MRI). Conventionally, this process is carried out manually in clinical practice but it is time consuming and observer dependent. It is a big challenge to segment cartilage manually from MRI, as cartilage structure has inadequate image contrast and complex tissue structure. Currently, semiautomatic and automatic methods are used to overcome this limitation. In the proposed method, we present a segmentation approach for the extraction of bone and cartilage from MRI. Here, in this paper we perform preprocessing for image enhancement and noise removal of knee MR image using Gaussian blur and block matching 3D method. Larger bones in the knee joint are segmented first to perform reliable cartilage segmentation. Hence distance regularization level set evolution (DRLSE) is used for bone extraction successively followed by cartilage segmentation. Further, the performance of the proposed method is analyzed on a variety of knee MR images and experimental results demonstrated the improvement in accuracy compared to existing approaches. In conclusion, it is observed that the proposed technique improves significant performance with consistency and robustness during the segmentation process. © Springer Nature Singapore Pte Ltd. 2020.