Face Detection Using Color Thresholding and Eigenimage Template Matching Diederik Marius Sumita...
-
Upload
spencer-shaw -
Category
Documents
-
view
219 -
download
0
description
Transcript of Face Detection Using Color Thresholding and Eigenimage Template Matching Diederik Marius Sumita...
Face Detection Using Color Thresholding and
Eigenimage Template Matching
Diederik MariusSumita Pennathur
Klint Rose
Approach
Input Image
YCbCrThresholding
Binary ImageProcessing
Separationof “Blobs”
SizeThreshold
Homogeneity/Aspect Ratio
ThresholdCross-covariancewith Eigenimage
DuplicateDetectionRemoval
Rejection
Match Face Location to Input Image
Indi
vidu
al b
lobs
x10
Skin Segmentation Used distribution of Cr and Cb
values for faces vs. background to thresholdSkin region: 105<Cr<135
140<Cb<165
Binary Image Processing Erosion and dilation with face-shaped
segmentation element Removes small foreground and
background objects Delineates between larger regions
Separation and Rejection Larger regions labeled and
separated Series of Rejection Thresholding:
Size of regions (>3000 total pixels) Aspect Ratio (0.5 <AR< 1.8 ) St. Dev. of image values (60<<100)
Size based thresholding
Coupled AR and St. Dev. based thresholdin
g
Eigenimages Computed from set of “good” faces Employed Sirovich-Kirby method to calculate first 10 eigenimages Eigenimage #2: most accurate location of center of face Used eigenimage #2 exclusively
1 2 3 4 5 6 7 8 9 10
Template Matching Found cross-covariance peak and
marked as potential center Removed face sized area from
image Found new highest peak - repeated
10x
False Detection Removal Thresholded detected peaks
to obtain potential faces only Determined whether
multiple peaks belong to the same person (neck, etc)
Rejection criteria: Near the same y-axis and within a predetermined vertical distance of a previous point
Rejectedpoints
Typical Detection Result Detection problems: rotated faces,
small faces, hands, faces in lower 1/3
Results Average run time: 100 seconds 95% of faces found 4.4% false positives
# Faces # Correct False Pos False Neg % Correct % FalseTraining Image 1 21 21 1 0 100 4.8Training Image 2 23 21 5 2 91 21.7Training Image 3 23 22 0 1 96 0.0Training Image 4 24 22 0 2 92 0.0Training Image 5 24 23 0 1 96 0.0Training Image 6 24 23 0 1 96 0.0Training Image 7 22 19 1 3 86 4.5Total 161 151 7 10 95 4.4
17
16
5 (0)4 (19)15
5 (0)4 (19)14
5 (0)4 (19)13
5 (0)4 (19)12
5 (0)15 (08)11
2 (1)2 (20)10
2 (1)1 (22)9
1 (2)13 (14)8
5 (0)12 (16)7
5 (0)11 (17)6
5 (0)14 (13)5
5 (0)9 (18)4
5 (0)9 (18)3
5 (0)2 (20)2
2 (1)4 (19)1
Gender RecognitionFace Detection
17
16
5 (0)4 (19)15
5 (0)4 (19)14
5 (0)4 (19)13
5 (0)4 (19)12
5 (0)15 (08)11
2 (1)2 (20)10
2 (1)1 (22)9
1 (2)13 (14)8
5 (0)12 (16)7
5 (0)11 (17)6
5 (0)14 (13)5
5 (0)9 (18)4
5 (0)9 (18)3
5 (0)2 (20)2
2 (1)4 (19)1
Gender RecognitionFace Detection