Facial Symmetry Analysis

11
Jason Mak

description

Jason Mak. Facial Symmetry Analysis. Project Statement and Motivation. The goal of this project is to automatically detect the facial symmetry axis and measure the symmetry of a given face. - PowerPoint PPT Presentation

Transcript of Facial Symmetry Analysis

Page 1: Facial Symmetry Analysis

Jason Mak

Page 2: Facial Symmetry Analysis

The goal of this project is to automatically detect the facial symmetry axis and measure the symmetry of a given face.

The purpose of this project is to create a form of biometric data, which could be useful for various detection processing or medical analysis.

Page 3: Facial Symmetry Analysis

1.)Skin Detection Use a Gaussian distribution of chroma (YCbCr) values from a

collection of skin samples, to detect possible skin candidates Apply an adaptive threshold to create a binary image of the

skin region

2.)Symmetry Axis Determination Use the binary image to determine the center of mass From the center of mass, crop out face region to remove

noise

3.)Gray Level Difference Calculation Compare the right side values to the left side values

correspondingly, pixel by pixel

4.) Misalignment Adjustment Shift symmetry axis and compare gray level difference, in

case symmetry axis is misaligned

Page 4: Facial Symmetry Analysis
Page 5: Facial Symmetry Analysis
Page 6: Facial Symmetry Analysis
Page 7: Facial Symmetry Analysis
Page 8: Facial Symmetry Analysis
Page 9: Facial Symmetry Analysis
Page 10: Facial Symmetry Analysis

Skin detection was a difficulty because of the small range in the database I chose. Improvement would come with a more thorough database of skin samples and a more accurate Gaussian distribution.

Symmetry axis determination was a difficulty because of the slight variations that came with each image. Although I was able to deal with misaligned axis, I could not account for images with tilted faces.

Skin detection was also a difficulty because lighting would effect the detection, causing skin regions to be neglected and non-skin regions to be accepted. Being able to account for lighting and texture differences would greatly improve such problems.

Page 11: Facial Symmetry Analysis