1977 年 33 巻 2 号 p. 147-155
70mm voiding cystogram is digitized row by row and punched out in the form of paper tape using JEC-5 small computer system and a film scanner. These digitized date are read into core memory in a 64×64 array, and written on a magnetic tape as one data per word using DP-5000 image processing system. This system is constructed of central processor TEAC-16 which has a performance of 16 kW core memory. Before extracting the boundary of urinary bladder and urethra, authors preprocessed following methods. (1) Correcting the output voltage to the relative exposure using a least squares method. (2) Eliminating the noise which is caussed by the film scanner using the digital filtering techniques. A crucial problem in using image segmentation method based on the thresholding is that of how to select the threshold. In many cases it is not possible to do this in advance, since the correct threshold may not be same for all voiding cystograms. However, it is sometimes possible to select a good threshold for each image automatically by examining the frequency distribution of gray levels.The frequency of-gray levels is constructed of the moving matrix which moves along element by element including the urinary bladder. In order to extract the urinary bladder and the urethra easily, authors reverse the digitized voiding cystogram. In the reversal image, as the urinary bladder and urethra have a maximum gray level, authors could segment the image by choosing the threshold at the 1st bottom of the valley bordering the 1st peak and the 2nd peak. In making the frequency distribution of gray levels, authors used 53×53 for the moving matrix size and used 50 for the number of gray levels.