Intravascular ultrasound (IVUS) imaging is an invasive technique based on catheter which has been attracted researcher’s attentions in the field of coronary artery disease. IVUS has many advantages over traditional methods of imaging coronary artery disease like coronary angiography: determination type, size and shape of plaque, the wall of arteries, … . In this paper, we used level set method (LSM) to segment lumen and non-lumen pixels and Hidden Markov Random Field (HMRF) to compute boundaries of lumen. This proposed method was evaluated on IVUS images of 7 patients. Our results had shown that using LSM-HMRF method, detection accuracy has increased up to 85%. Results shown that combination of LSM-HMRF could successfully identify the lumen boundary. The advantage of this method is that we obtained one pattern using LSM from all of IVUS images that have. And by this one we searched for similarity among other IVUS images to perform segmentation. HMRF method do this search. And since we had one pattern the computational load decreased. As we know HMRF method is based search of graph. We proposed to divide images into 4 part and then search for similarity between obtained pattern and IVUS images. And because of this modification in HMRF method we were more precise. One can say why we used only one pattern for all IVUS images. Our answer is that we obtained a general pattern which have common features from all images. So it is representative of all.