Face detection and recognition depend strongly on illumination conditions. In this paper, we present improvements in
two illumination compensation methods for face recognition. Using genetic algorithms (GA) we select parameters of the
Discrete Cosine Transform (DCT) and Local Normalization (LN) methods to improve face recognition. In the DCT
method all low frequency components within an isosceles triangle, of side Ddis, are eliminated. The best results were
reported for Ddis=20. In the LN method it is proposed to normalize the value within a window by the mean and standard
deviation. Best results were reported for window sizes of 7x7. In the case of the DCT method, we assigned weights to
eliminate the coefficients of the low frequency components using a GA. In the case of the LN method for a fixed
window size of 7x7, we selected the normalization method by a GA. We compare results of our proposed method to
those with no illumination compensation and to those previously published for DCT and LN methods. We use three
internationally available face databases Yale B, CMU PIE and FERET where the first two contain face images with
significant changes in illumination conditions. We used Yale B for training and CMU PIE and FERET for testing. Our
results show significant improvements in face recognition in the testing database. Our method performs similarly
or slightly better than DCT or LN methods in images with non-homogeneous illumination and much better than DCT or
LN in images with homogeneous illumination.
Real-time face detection on video sequences is important in diverse applications such as,
man-machine interfaces, face recognition, security and multimedia retrieval. In this work,
we present a new method based on the maximization of local components in the directional
image to optimize templates for frontal face detection. In the past, several methods for face
detection have been developed using face templates. These templates are based on common
face features such as eyebrows, eyes, nose and mouth. Templates have been applied to a
directional image containing faces computing a line integral to detect faces with high
accuracy. In this paper, the maximization of local components in the directional image is
used to select new templates optimizing its size and response to a face in the directional
image. The method selects common directional vectors in a set of frontal faces to generate
the template. The method was tested on 386 images from the Caltech face database and 55
images from the Purdue database. Results were compared to those of the traditional
anthropometric templates that contain features from the eyebrow, nose and mouth. Results
show that the new templates have significant better performance in the estimation of face
size and the line integral value. Face detection reached 97% on the Caltech face database
and 98% on the Purdue database. The templates have fewer number of points compared to
the traditional anthropometric templates which will lead to lower processing time.
Real-time face and iris detection on video sequences is important in diverse applications such as, study of the eye function, drowsiness detection, man-machine interfaces, face recognition, security and multimedia retrieval. In this work we present and extension to our previous method to incorporate face and iris detection in faces with coronal and transversal axis rotations in real time. The method is based on anthropometric templates and consists of three stages: coarse face detection, fine face detection and iris detection. In the coarse face detection, a directional image is computed and the contribution of each directional vector is weighted into an accumulator. The highest score in the accumulator is taken as the coarse face position. Then, a high-resolution directional image is computed. Face templates were constructed off-line for face coronal and transversal rotation, using face features such as elliptical shape, location of the eyebrows, nose and lips. A line integral is computed using these templates over the fine directional image to
find the actual face location, size and rotation angle. This information provides a region to search for the eyes and the iris boundary is detected within this region by a ratio among to line integrals using a semicircular template. Results computed on five video sequences which include coronal and transversal rotations with over 1900 frames show correct face detection rate above 92% and iris detection rate above 86%.
Real-time face and iris detection on video images has gained renewed attention because of multiple possible applications in studying eye function, drowsiness detection, virtual keyboard interfaces, face recognition, video processing and multimedia retrieval. In this paper, a study is presented on using directional templates in the detection of faces rotated in the coronal axis. The templates are built by extracting the directional image information from the regions of the eyes, nose and mouth. The face position is determined by computing a line integral using the templates over the face directional image. The line integral reaches a maximum when it coincides with the face position. It is shown an improvement in localization selectivity by the increased value in the line integral computed with the directional template. Besides, improvements in the line integral value for face size and face rotation angle was also found through the computation of the line integral using the directional template. Based on these results the new templates should improve selectivity and hence provide the means to restrict computations to a fewer number of templates and restrict the region of search during the face and eye tracking procedure. The proposed method is real time, completely non invasive and was applied with no background limitation and normal illumination conditions in an indoor environment.
Real-time face and iris detection on video sequences has been used to study the eye function and in diverse applications such as drowsiness detection, virtual keyboard interfaces, face recognition and multimedia retrieval. A non-invasive real time iris detection method was developed and consists of three stages: coarse face detection, fine face detection and iris detection. Anthropometric templates are used in these three stages. Elliptical templates are used to locate the coarse face center. A set of anthropometric templates which are probabilistic maps for the eyebrows, nose and mouth are used to perform the fine face detection. Face rotations are considered by rotating the anthropometric templates in fixed angles in steps of 10 degrees. Iris position is then determined within the eye region using another template with concentric semi-circles to compute a line integral in the boundary iris-sclera. The position with the maximum value indicates the iris center. The new method was applied on 10 video sequences, with a total of 6470 frames, from different people rotating their faces in the coronal axis. Results of correct face detection on 8 video sequences was 100%, one reached 99.9% and one 98.2%. Results on correct iris detection are above 96% in 9 of the video sequences and one reached 77.8%. The method was implemented in real-time (30 frames per second) with a PC 1.8 GHz.
Proceedings Volume Editor (4)
This will count as one of your downloads.
You will have access to both the presentation and article (if available).
Access to the requested content is limited to institutions that have purchased or subscribe to SPIE eBooks.
You are receiving this notice because your organization may not have SPIE eBooks access.*
*Shibboleth/Open Athens users─please
sign in
to access your institution's subscriptions.
To obtain this item, you may purchase the complete book in print or electronic format on
SPIE.org.
INSTITUTIONAL Select your institution to access the SPIE Digital Library.
PERSONAL Sign in with your SPIE account to access your personal subscriptions or to use specific features such as save to my library, sign up for alerts, save searches, etc.