Face Detection

							Smart Traveller with Visual
       Translator
What is Smart Traveller?

• Mobile Device which is convenience for a
  traveller to carry
• E.g. Pocket PC, Mobile Phone
What is Visual Translator?

• Recognize the foreign text and translate it
  into native language
• Detect the face and recognize it into name
Requirements

• Simple (Computational low power)
• Lightweight (Low Storage)
• User Friendly
     Core Pattern Recognition Model

              Find Each                Quantify the                  Assign Label
              Object from              object by some                for each object
              the Image                characteristics
                 Image                       Feature                  Classification
              Segmentation                  Extraction
Input Image                  Object Image                Feature Vector                Object Type
Character Recognition

• Language: Korean
• Target: Sign, Guidepost
  – Contrast in Color
  – Printed Character
Image Segmentation
• Binarization
  – Using Color Histogram to binarize the image for the
    background and the character
• Text Region Segmentation
  – User Define Method
  – Edge Detection with horizontal and vertical
    projections
• Stroke Extraction
  – Labeling of connected component Algorithm
Feature Extraction

• Stroke Features
  – Number of Junctions, Corners
  – Any Hole
• Gabor Features
Recognition

• Minimum Euclid Distance
• Learn the Decision Tree by training
 examples
Demo
          Face Detection
Outline
•Find Face Region
•Find the potential eye region
•Locate the iris and eyelids
Find Face Region - Color-based
model
• We used this method because of its
  simplicity and robustness.
• Usually RGB color model will be
  transformed to other color modes such as
  YUV (luminance-chrominance) and HSB
  (hue, saturation and brightness)
YUV

• We use YUV or YCbCr color model.
• Y component is used to represent the
  intensity of the image
• Cb and Cr are used to represent the blue
  and red component respectively.
YCbCr Image
• Y, Cb ,Cr component image




        Y            Cb       Cr
Representation of skin color
• We just use a
  simple ellipse
  equation to model
  skin color.

                  Cr




                       Cb
Representation of skin color




 • The white regions represent the
   skin color pixels
Color segmentation
• We distribute some agent in the image uniformly.
• Then each agent will check whether the pixel is
    a skin-like pixel and not visited by the other
    agent.
•   If yes, it will produce 4 more agents at its four
    neighboring points.
•   If no, it will move to one of four neighboring
    points randomly and decrease its lifespan by 1.
    When its lifespan becomes zero, it will be
    removed from the image.
    Color segmentation
•   This agent produce 4
    more agents
Color segmentation

• The advantage of this algorithm is that we
  need not to search the whole image.
• Therefore, it is fast.
Color segmentation
• 19270 of 102900
    pixels is searched
    (about 18.7%)
•   There are 37 regions
•   Each color regions
    represent each
    regions searched by a
    father agent
Eye detection
• After the segmentation of face region, we
  have some parts which are not regarded
  as skin color.
• They are probably the region of eye and
  mouth
• We only consider the red component of
  these regions because it usually includes
  the most information about faces.
Eye detection
• We extraction such
    regions.
•   The red region
    represent the region
    which is not skin color.
Eye detection

We do the following on the regions of
   potential eye region
1. Histogram equalization
2. Threshold
3. Template matching
Eye detection
  Histogram equalization




  Threshold with < 49



  Template Matching
Locating the iris and eyelids
We plan to use the following methods to
 improve the face detection
We can use these methods to locate the iris
 and eyelid precisely.

Template matching
  – Correlation variance filter
  – Deformable template
END