HEATMAP ANALYSIS OF PROFILE FACIAL IMAGES

Facial images are a very common element on the World Wide Web and in mobile applications. With the widespread use of eye-tracking systems, research in the field of viewing and memorizing facial images has greatly expanded. The main issue in analyzing facial images is to understand the way of viewing, memorizing and recognizing. Most of the research is done for the frontal facial images, but we focused on profile facial images. One of the most widely used methods for determining memorization and recognition of facial images is the internal facial feature method. However, in our research we have focused on the new heatmap method. In this method, the results of area, perimeter and circularity of heatmaps are analyzed. For the whole research we used a standardized memory test comprising two phases: an observation test and a recognition test. To analyze the influence of time on recognition success, we set four display times of facial images in observation test (1s, 2s, 4s, 8s). The results show that the recognition success increases with increasing display time of the facial images in the observation test. The turning point occurs at a display time of 4 seconds, as the recognition performance in the 8s test is almost the same as in the 4s test. The heatmap analysis method shows an increase in the area and perimeter of the heatmap areas up to the 4s test, after which both values remain practically the same. After that, both values remain practically the same. At this point, the circularity of the viewing areas also no longer decreases and is the same in the 8s test as in the 4s test. The results of the heatmap analysis are therefore in good correlation with the results of the facial recognition and proved the usefulness of heatmap analysis for research into the observation and recognition of facial images.