In medical X-ray examinations, X-ray collimators are frequently employed to shield the body parts from unnecessary X-ray exposure and to improve the image quality. The confined regions (collimations) are not regarded as diagnostically important. And it has been pointed out that the presence of collimations may impair diagnosis of subtle lesions due to dazzle. For efficient image compression, it is desirable that the collimations are accurately detected. The purpose of this study is to develop a computerized method to correctly identify collimations in digital radiographs. In this study, we propose a method of applying plane detection Hough transform where digital radiographs are regarded as three-dimensional data by adding one-dimension of the brightness direction, in order to detect the boundaries between collimations and irradiation fields as planes. Therefore, we describe a validity of the plane detection by modeling the collimation boundaries. In addition, the effectiveness in compressing the data of a processed images greater than the original ones.