Kwon and Lobo, one of the first papers related to age estimation from digital images of the face, developed an algorithm based on ratios of different facial features and a wrinkle analysis, including the automatic extraction of the required features. The facial images are then classified into three age groups: babies, young adults, and senior adults. Their experiment resulted in 100\% accuracy.
However, their database was limited to 47 high resolution photos with that would be very difficult to implement in a practical application. Horng et al. recognized these limitations and used wrinkle and geometric features to classify four age groups, using the Sobel filter to judge the degree of wrinkles. To determine the wrinkle features, the density and depth of the wrinkle and the average variance of the skin were extracted, using the Sobel edge magnitude. Hayashi et al. studied the pattern of face wrinkles from a database of controlled face images taken from 300 of subjects ranging from 15 to 64 years of age for estimating age and gender. Their approach involved the extraction of skin regions for the purpose of quantifying and identifying the types of wrinkles, short or long. The wrinkles were enhanced by histogram equalization on the raw skin regions and the wrinkle classification was performed by the Digital Template Hough Transform.