Estimation of the nutritional balance of food dishes by color analysis of photographs - using salad as a model dish

抄録

The possibility of the estimation of nutritional balance using color analysis of photographs of food dishes was investigated in the case of salad. A total of 78 photographs of different salad varieties obtained from a recipe book and recipe website were analyzed. The ingredients of each salad were painted with 10 different colors and the area % of each color was calculated using the Feelimage Analyzer (VIVA Computer Ltd.). The weight average food composition was determined for each color using recipes indicating trace nutrients and dietary fiber. The amounts of ingredients estimated by color analyses were corrected using correlation equations obtained from the correlation analyses between the amount of nutrients estimated from the area % of the color and the amounts calculated per 100 g of salad from the recipe. The estimated nutrient content obtained using the weight average food composition for each color was then compared with those calculated per 100 g of salad based on the recipe. Of the 78 varieties of salad analyzed, in more than 50 % of cases the contents of potassium, dietary fiber, vitamin B₁ and phosphorus could be estimated to within ±30 % of those calculated from recipe. In order to validate this method, an additional three varieties of salad were analyzed. While the possibility of the evaluation of nutritional balance using color analysis of salad photographs was suggested by the results of this study, an increase in accuracy to estimate the amounts of ingredients and the weight average food composition for each color will be required.