Presentation is loading. Please wait.

Presentation is loading. Please wait.

Face Detection – EE368 Group 10 May 30, 2003 1 Face Detection EE 368 Group 10 Waqar Mohsin Noman Ahmed Chung-Tse Mar.

Similar presentations


Presentation on theme: "Face Detection – EE368 Group 10 May 30, 2003 1 Face Detection EE 368 Group 10 Waqar Mohsin Noman Ahmed Chung-Tse Mar."— Presentation transcript:

1 Face Detection – EE368 Group 10 May 30, 2003 1 Face Detection EE 368 Group 10 Waqar Mohsin Noman Ahmed Chung-Tse Mar

2 Face Detection – EE368 Group 10 May 30, 2003 2 Overview  Project goals  Implementation  Skin color segmentation  Morphological processing  Connected region analysis  Template matching  Female recognition  Results

3 Face Detection – EE368 Group 10 May 30, 2003 3 Project Goal  Detect and locate human faces in color images similar to those from the training set  Limited variations in zoom and lighting  Similar scene conditions  Many occlusions

4 Face Detection – EE368 Group 10 May 30, 2003 4 Implementation Overview Skin Color Segmentation Morphological Processing Connected Region Analysis Template Matching Face Coordinates Input Image

5 Face Detection – EE368 Group 10 May 30, 2003 5 Skin Color Segmentation Hue Saturation

6 Face Detection – EE368 Group 10 May 30, 2003 6 Skin Color Segmentation  Non-skin color regions have been eliminated  Arms, hands, and various regions still remain

7 Face Detection – EE368 Group 10 May 30, 2003 7 Morphological Processing Convert to Grayscale Intensity Thresholding Morphological Opening Fill Holes Mask Image Color-segmented Image Morphological Opening I = rgb2gray(colorSegImage) I(find(I<=50))=0 se=strel(‘disk’,1) I=imopen(I,se) I=imfill(I,’holes’) se=strel(‘disk’,6) I=imopen(I,se)

8 Face Detection – EE368 Group 10 May 30, 2003 8 Morphological Processing Before After

9 Face Detection – EE368 Group 10 May 30, 2003 9 Connected Region Analysis (Geometry)  Reject regions that are  Narrow  Short  Narrow and tall  Wide and short  Thresholds derived from training image statistics

10 Face Detection – EE368 Group 10 May 30, 2003 10 Connected Region Analysis (Euler Number) Reject regions with Euler number greater or equal to 0

11 Face Detection – EE368 Group 10 May 30, 2003 11 Result of Connected Region Analysis Before After

12 Face Detection – EE368 Group 10 May 30, 2003 12 Template Matching

13 Face Detection – EE368 Group 10 May 30, 2003 13 Template Matching

14 Face Detection – EE368 Group 10 May 30, 2003 14 Female Face Detection  Find the face with the white scarf  Draw a box around the face centroid and count the number of white pixels  Find the face with the long hair  Draw a box around the face centroid and count the number of black pixels  The one with the largest number of black or white pixels is designated as a female face

15 Face Detection – EE368 Group 10 May 30, 2003 15 Results Training ImageTotal facesDetectedFalse PositiveRepeat Hit 121 00 2242310 3252400 4 00 5 2200 624 00 722 00 TOTAL16416010 Detects 160 out of 164 faces with 1 false positive. Average run- time on a Pentium 4 1.8GHz PC is 35 seconds.

16 Face Detection – EE368 Group 10 May 30, 2003 16 Conclusion  Face detection program with 97% accuracy over the training images  Run-time under a minute  Hardest part of project is separating the connected faces  Solved with successive template matching and blacking out face regions  Not very robust, sensitive to scene conditions in the images

17 Face Detection – EE368 Group 10 May 30, 2003 17

18 Face Detection – EE368 Group 10 May 30, 2003 18


Download ppt "Face Detection – EE368 Group 10 May 30, 2003 1 Face Detection EE 368 Group 10 Waqar Mohsin Noman Ahmed Chung-Tse Mar."

Similar presentations


Ads by Google