Arterial spin labeling (ASL) allows non-invasive imaging and quantification of brain perfusion by magnetically labeling blood in the brain-feeding arteries. ASL has been used to study cerebrovascular diseases, brain tumors and neurodegenerative disorders as well as for functional imaging. The use of a perfusion template could be of great interest to study inter-subject regional variation of perfusion and to perform automatic detection of individual perfusion abnormalities. However, low spatial resolution and partial volume effects (PVE) issues inherent to ASL acquisitions remain to be solved. The purpose of this study is to enhance the template quality by using DARTEL non-rigid registration and by correcting for PVE. PICORE-Q2TIPS ASL datasets were acquired on 25 healthy volunteers at 3T. Four methods of creating the template were evaluated using leave-one-out cross correlation. Subsequently, these methods were applied to hyper-perfusion detection on functional ASL data of 8 healthy volunteers and compared with the standard generalized linear model (GLM) activation detection.