Interactive Feature Visualization and Detection for 3D Face Classification

Interactive Feature Visualization and Detection for 3D Face Classification

Jason McLaughlin (Indiana University-Purdue University, USA), Shiaofen Fang (Indiana University-Purdue University, USA), Sandra W. Jacobson (Wayne State University, USA and University of Cape Town, South Africa), H. Eugene Hoyme (Sanford School of Medicine, USA), Luther Robinson (State University of New York, USA) and Tatiana Foroud (Indiana University, USA)
DOI: 10.4018/jcini.2011040101
OnDemand PDF Download:
No Current Special Offers


A new visual approach to the surface shape analysis and classification of 3D facial images is presented. It allows the users to visually explore the natural patterns and geometric features of 3D facial scans to provide decision-making information for face classification which can be used for the diagnosis of diseases that exhibit facial characteristics. Using surface feature analysis under a digital geometry analysis framework, the method employs an interactive feature visualization technique that allows interactive definition, modification and exploration of facial features to provide the best discriminatory power for a given classification problem. OpenGL based surface shading and interactive lighting are employed to generate visual maps of discriminatory features to visually represent the salient differences between labeled classes. This technique will be applied to a medical diagnosis application for Fetal Alcohol Syndrome (FAS) which is known to exhibit certain facial patterns.
Article Preview


In recent years, interactive visualization and 3D computer graphics techniques have started to play significant roles in data mining and data analysis applications. The feedback and involvement of human intuition and knowledge through graphical interfaces and interactive visualization can often lead to faster and more reliable data mining and analysis outcomes. This visual analytics approach is particularly effective for the analysis of visual/spatial data from 3D sensors, as spatial data have natural visual representations. On the other hand, the rapid advances of sensory technologies have led to the enormous proliferation of multimedia data and have created a great challenge and an urgent need for new and more effective data mining and analysis techniques for data from multimedia sources. Many applications, such as medical diagnosis and biometrics, can benefit greatly from more reliable and efficient 3D data analysis methods and techniques.

This paper presents a new visual approach to the analysis of polygon mesh surfaces of human faces for a medical diagnosis application: Fetal Alcohol Syndrome (FAS). FAS is a neurological disorder resulting from prenatal exposure to alcohol. It is the most common nonhereditary cause of mental retardation and is often associated with growth deficiencies and developmental abnormalities of the central nervous system, and a pattern of various facial anomalies (Moore, Ward, Jamison, Morris, Bader, & Hall, 2002). It is estimated that the prevalence of FAS in the general population is likely to be between 0.5 and 2.0 per 1,000 births (May & Gossage, 2001). While there are several FAS diagnostic criteria used in the medical community (Jones & Smith, 1973; Moore, Ward, Jamison, Morris, Bader, & Hall, 2002; Jacobson, 2008), identification of the distinctive pattern of facial features anomalies is a key and necessary component for the diagnosis and the only one at this time that can be readily detected and determined and potentially automated.

In this study, 3D facial images are collected using an eye-safe laser scanner. Multiple scans are made and then stitched together to form a dense and irregular polygon mesh surface. The analysis of polygon mesh surfaces is a new research approach in medical imaging, as existing 3D medical imaging technologies focus on volumetric images, such as CT and MRI scans. There is, however, a tremendous need in medical applications for the image analysis of surface scans that capture detailed surface geometry and texture information. For some neurological disorders, facial feature anomalies are highly characteristic and unique. 3D data analysis techniques that capture the correlation between face and disease can potentially provide effective diagnosis tools for medical research and clinical studies, especially in the pre-screening process and in early diagnosis of children. It has also been shown in Jain and Hoffman (2005) that the traditional 2D facial analysis approach is not effective for this type of diagnosis problems, since the features affecting the facial dysmorphology are highly three-dimensional and cannot be easily captured by 2D photos.

Complete Article List

Search this Journal:
Volume 16: 1 Issue (2022)
Volume 15: 4 Issues (2021)
Volume 14: 4 Issues (2020)
Volume 13: 4 Issues (2019)
Volume 12: 4 Issues (2018)
Volume 11: 4 Issues (2017)
Volume 10: 4 Issues (2016)
Volume 9: 4 Issues (2015)
Volume 8: 4 Issues (2014)
Volume 7: 4 Issues (2013)
Volume 6: 4 Issues (2012)
Volume 5: 4 Issues (2011)
Volume 4: 4 Issues (2010)
Volume 3: 4 Issues (2009)
Volume 2: 4 Issues (2008)
Volume 1: 4 Issues (2007)
View Complete Journal Contents Listing