Automatic human ear detection approach using modified adaptive search window technique

Abstract

The human ear biometric recognition plays an important role in the forensics specialty and has significant impact for biometrician scientists and researchers. Actually, many ear recognition researches showed promised results, but some issues such as manual detection process, efficiency and robustness aren't attained a certain level of maturity. Therefore, the enhancement developing approaches still continuous to achieve limited successes. We propose an efficient, reliable and simple automatic human ear detection approach. This approach implement two stages: preprocessing and ear landmarks detection. We utilized the image contrast, Laplace filter and Gaussian blurring techniques to made enhancement on all images (increasing the contrast, reduce the noisy and smoothing processes). After that, we highlighted the ear edges by using the Sobel edge detector and determining the only white pixels of ear edges by applying the image substation method. The improvement focused on using the modified adaptive search window (ASW) to detect the ear region. Furthermore, our approach is tested on Indian Institute of Technology (IIT) Delhi standard ear biometric public dataset. Experimental results presented a well average detection rate 96% for 493 image samples from 125 persons and computational time almost ≈ 0.485 seconds which is evaluated with other previous works.