- identify a face from a database of faces
- determine whether an image contains a face present in the database
- to decide whether an image represents a face or not
We will be using the Yale Faces Database.
This database contains 120 greyscale images, representing the faces of 15 people. There are 8 images per person.
The method developed by Turk and Pentland defines the eigenfaces as the main axes obtained by carrying out a principal component analysis (PCA) of the vectors associated with the reference faces.
We will use singular value decomposition (SVD).
example of reconstruction of an image the split and merge algorithm is used for image segmentation.
it works like this:
- Split the image into equal size regions
- Calculate homogeneity for each region
- If the region is homogeneous, then merge it with neighbors
- The process is repeated until all regions pass the homogeneity test
We will compute a measure of similarity between two images from the normalized histogram. This measure of similarity will be used in order to find images that are the most similar to a given image.
20 best matches of image 'Liontigre1.png' 
The goal of this practial work is to experiment various edge detectors.
- comparison between the first and second order detectors
- study of the impact of smoothing
- removing non maxima answers of the detectors
- evaluation in term of robustness and localization
