Figure 1
From: Towards joint segmentation and registration of the myocardium in CP-BOLD MRI at rest
The figure at top shows the flowchart of the algorithm. First 9 × 9 patches are extracted for each pixel and concatenated with HOG and Gabor features of each patch. Dictionaries DB and DM are calculated for the initialization of registration process. The sparse representation of each pixel p for background XB and myocardium XM are calculated and concatenated; Xp =[XB, XM ]. In data term of the objective function of registration; the similarity metric in between two consecutive frames is defined as L1 norm of the concatenated supports: S(It(p), It+1(p+u)) = ||Xtp-Xt+1p+u||1. The approach iterates refinement step of the segmentation using the registration for the agreement of classification and then finding the D, then X. The fusion of labels in between the different frames is used for classification of the myocardium region.