In this paper we present a learning framework for segmentation and tracking in 2D cardiac tagged MRI sequences. We employ a transformed component analysis (TCA) algorithm to estimate the shape variations, and at the same time, eliminate the rotation distortions of the training shapes. This method also integrates the motion and the static local appearance features and generates accurate boundary criteria via a boosting approach. We extend the conventional Adaboost classifier into a posterior probability form, which can be embedded in a particle filter based shape tracking framework. The TCA shape representation is used to constrain the shape variations and lower the dimensionality, so that it makes the tracking process more robust and faster. We also learn two shape dynamic models for systole and diastole separately to predict the shape evolution. Our segmentation and tracking method incorporates the static appearance, the motion appearance, the shape constraints, and the dynamic prediction in a unified way. The proposed method has been applied to 50 tagged MRI sequences. The experimental results show the accuracy and robustness of our approach.