We estimated nutritional intake from color photographs taken of meals with a 35-mm camera, and compared them with the values calculated from the menu. Among 21 meals examined, the coefficient of correlation between the values by the photo method and those from examining the menu was appreciably high; 0.842 in energy, 0.828-0.931 in three major nutrients, and 0.745-0.940 in vitamins and minerals. The differences between the photo method and menu observation were 0-14.8% on average, and the coefficient of variation of the differences was 9.1-32.3%. Estimation of salt was, however, very poor, and the agreement was lower due to the cooking method used. Some differences were found among the dietitians in estimated values of several nutrients. The running cost was approximately 87-yen per meal, which may pose a problem for large-scale dietary surveys. Compared with the use of record, recall, and frequency questionnaire methods, the photo method was easily used by participants, especially those not used to cooking, and could be applied to eating out or for take-out food. Moreover, this method may provide not only quantitative data but also qualitative including eating atmosphere. When estimation is difficult due to the ingredients or recipe, combined use with record or recall methods can facilitate estimation. This photo method is thus considered to be a valid and feasible means of dietary assessment.