This study proposed a method of nose and nostril detection in thermal images using the Viola-Jones method. Since thermal images show facial features with unclear boundaries, using the Viola-Jones method to identify features results in hits and false alarms. To eliminate false alarms, the face was first detected by setting size and temperature constraints. Next, location constraints were added to precisely locate the nose and nose holes. After detecting the nose and nose holes, template matching

more »

... was used to track the nose holes, and breathing was monitored by calculating the average temperature in the nostril region. When the proposed method was applied to thermal images captured for one minute each from 20 subjects, the nose and nose holes of all 20 subjects were detected with 100% accuracy within 3 seconds. The nose holes were also successfully tracked for all cases. Breathing rates were compared for measurements obtained from average nostril temperature and from a monitoring device. This revealed a match for 16 out of the 20 subjects, and only one different measurement for 2 subjects.