Intra-cardiac masses identification in echocardiograms is an important task in cardiac disease diagnosis. To improve the diagnostic accuracy, the tumor and thrombi is identified using an automatic classification method based on the sparse representation. A region of interest is cropped to define the mass area. Then, a globally demising method is employed to remove the speckle and preserve the structures. Subsequently, the contour of the mass and its connected aerial wall are described by our proposed adaptive co segmentation localized region level set model. These methods are used to distinguish intra cardiac tumor and thrombi in echocardiography.