要約

The use of high-frame-rate neutron radiography has been proposed as a method for recording images of oil in a running car engine. However, it is difficult to capture the behavior of oil from such images, because the images are dark and blurry. Therefore, we attempt to overcome this problem using a multi-scale image processing technique, in which the oil regions are classified based on scale. This method can be used to focus attention on objects with a particular scale. In addition, we propose a technique by which to capture the time-dependent behavior of oil using spatio-temporal image analysis. Using these techniques, the region of interest can be determined and observed carefully.