In this paper, we suggest an automatic fat region setting method in computed tomography to measure body fat. Quantitative measurement of body fat is important for the diagnosis and treatment of diseases related to obesity. The evaluation methods of obesity such as weight and height ratio, waiste and hip circumference ratio, and subcutaneous fat thickness by clipper, have a weak point that they do not measure body fat quantitatively. To overcome the weak point a body fat measurement method by computed tomography has been tried recently. The body fat measurement by computed tomography presents volume of body fat quantitatively at a specific region such as subcutaneous and abdominal fats. Computed tomography provides an image consisting of voxels with Hounsfield unit which is proportional to the density at the voxels. In general, the Hounsfield units of the body fat is known as it has the values from -150 to -50. However, the Hounsfield unit range of the fat is different for every person and every region of a person. In this paper, we show the difference of the body fat range in the Hounsfield units for 20 persons and various regions of a person, and we suggest an automatic setting method of the body fat region in the Hounsfield units due to the Gaussian function fitting of the histogram in the computed tomography image.