Hippocampus segmentation is one of the significant medical imaging procedures for the early detection of Alzheimer's disease. A new approach is proposed to delineate the hippocampus from the Magnetic Resonance Images, which is pre-processed with the help of wave atom shrinkage. The robustness of this method is the inclusion of knowledge about the neuro-anatomy, radiological properties and variability of the hippocampus. Quantitative comparison showed a close correlation between the algorithm and the human expert as well as high spatial overlap between the ROI generated by the two methods, and it was able to identify the hippocampus under tissue variability.