Face and Full-body Image Processing
- Members: John Robinson, Andy Pomfret, Justen Hyde, Matt Day, Jie Li, Ziran Wu
- Website: http://www.intuac.com/userport/john/face.html
- Start Date: March 2005
A new scheme for covariance matrix regularization [1] gives excellent results when applied to appearance-based face processing. Hierarchical Covariance Estimation regularizes the plug-in estimates from statistics by mixing with assumed priors, global statistics and noise. The result is that conventional maximum likelihood classification yields results significantly superior than with unregularized models or with earlier covariance estimators. For example, for face/non-face classification, this graph shows how HCR error rates vary with training set size compared with Regularized Discriminant Analysis and Mixed LOOC1 regularization.
This means that ML face/non-face classification can be applied to face detection by running a window over a picture and finding matches. Some results are shown here:
On the bottom row the left picture includes a false positive and the right picture a false negative. The results are noteworthy not for being significantly better than the sophisticated face detectors now available, but for achieving comparable performance using the most conventional of classical models with well-regularized covariance estimates.
Using conditional density estimation, 3D face structure can be recovered from mugshots without explicit modeling of shape [2]. With regularized normal models, any missing measurements are estimated as the expected value of a conditional density given the known measurements. Having trained a model with greyscales plus depths, the greyscales of a new face image form a "probe" into the model to recover depths.
Similarly, the grey levels of a face can be used to estimate descriptive attributes such as gender, age, ethnicity and expression [3]. Alternatively for the same purpose of face description we can use cascade classifiers. Some of our work on these was described in [4].
We are applying similar techniques to the detection of complete humans and description of their pose and other attributes.
For more details about earlier activity in face image processing, see the summary here and related pages (follow the links under "Imaging Humans").
[1] J A Robinson, Covariance estimation in full- and reduced-dimensionality image classification, Image and Vision Computing, 27(8):1062-1071, 0262-8856, July, 2009.
[2] J A Robinson, J R Hyde, Estimation of face depths by Conditional Densities British Machine Vision Conference 2005, Vol 2, pp 609-618, Oxford, Sept. 2005. PDF
[3] J A Robinson, Regularized Single-Kernel Conditional Density Estimation for Face Description, IEEE International Conference on Image Processing (ICIP), Cairo, Egypt, November, 2009, pp 2461-2464.
[4] M G Day, J A Robinson, Constructing efficient cascade classifiers for object detection, IEEE International Conference on Image Processing (ICIP), Hong Kong, China, September 2010, pp 3781-3784.