We have studied methods to assess stress variability from real images. Stress variation can be estimated from the state of subcutaneous blood flow. To obtain the subcutaneous blood flow state from real images, we focused on the property that the absorptivity to the subcutaneous tissue differs depending on the color of the light. Based on this property, we thought that the R-B values could be used to extract the blood flow state. During the discomfort experiment, the R-B values in the nasal region showed correspondence with stress fluctuations, indicating its usefulness as a stress estimation method. However, skeletal differences and body movements change the position and size of light reflections and shadows in the real images. These effects of light intensity may have reduced the accuracy of stress assessment. As a method to remove the influence of light intensity, this study focuses on the relationship between color information and blood flow information for each pixel obtained from real images. Since color information differs from pixel to pixel, we introduce a weighting method that changes the degree of focus on a pixel according to the features indicating blood flow conditions extracted from the color information. By dividing the nasal region into multiple regions and introducing a weighting method for each region, we were able to classify both regions that are significantly affected by light intensity and regions that represent blood flow conditions. The proposed method made it possible to extract regions that contain only the blood flow state independent of light intensity.