Markov Random Field Image Segmentation Based On Graph Cut

Image segmentation plays a significant role in the field of computing vision,which is a key step in image processing to help researchers better understand images.Since image segmentation has been an independent research project,a variety of image segmentation algorithms also arise at the historic moment to make it more targeted to solve the existing problems.Among them,Markov Random Field Model(MRF)algorithm is an important research content in the Field of image segmentation in recent years.Its prototype is a probability and statistics mathematical model,and its concise and efficient algorithm process is popular with people.As the application range of the MRF model is increasing in the image segmentation fields,the traditional MRF model algorithm has single efficacy problems such as timing,accuracy,and other aspects of demand that can't be taken into account.So we made some improvements and combined it with Graph Cut(GC)algorithm to improve the segmentation efficiency.Here's our main researches:(1)When you segment an image,the traditional MRF model can't account for the interconnection between the pixels,which makes it less accurate.To solve these problems,the MRF linear variable weight image segmentation method has been proposed in this paper.First,it adds the intensity information between the neighboring pixels in the marked field and the characteristic field,so the image spatial information can be effectively used.Then,the exponential variable weight parameter will be changed to the linear variable weight parameter to connect the marker field and the characteristic field.Thus,it can accelerate the update speed of the segmentation result and increase the selection range of the potential function.The experiments show that the algorithm presented in this paper is more effective and robust in the accuracy and regional consistency of the segmentation results.Both in the segmentation speed and image processing efficiency,there has been great improvements.(2)When the MRF model was minimized with IRGC(Iteratively reweighted graph cut)algorithm,it is easy to fall into local minimum.In this paper,an IRGC-Swap algorithm under MRF model is proposed.IRGC-Swap algorithm introduces ?-? swap operator in the iterative process of IRGC.Thus,not only can we ensure the continuous energy optimization of MRF model by IRGC algorithm,but also we can improve the microcirculation,reduce iterative times,solve the problem of energy fallen into local minimum and improve search efficency through ?-? swap operator.To prove the validity of IRGC-Swap algorithm,plenty of experiments have been put in practice.And the results show that this improved algorithm can achieve a small amount of energy in a short period of time.Besides it optimizes better than any other MRF energy minimization method in existence.