Fast and Accurate Face Identification Using Overlapping DCT
In the JPEG image compression algorithm, the input image is divided into 8-by-8 or 16-by-16 blocks, and the two-dimensional DCT is computed for each block. The DCT coefficients are then quantized, coded, and transmitted. The JPEG receiver (or JPEG file reader) decodes the quantized DCT coefficients, computes the inverse two-dimensional DCT of each block, and then puts the blocks back together into a single image. For typical images, many of the DCT coefficients have values close to zero; these coefficients can be discarded without seriously affecting the quality of the reconstructed image. Such algorithm results particularly robust also for face identification. Moreover the 2D DCT operator can be applied to overlapping data.
The extracted feature vectors are used as input to a simple nearest neighbor algorithm. The k-nearest neighbor algorithm is amongst the simplest of all machine learning algorithms. An object is classified by a majority vote of its neighbors, with the object being assigned to the class most common amongst its k nearest neighbors. k is a positive integer, typically small. If k = 1, then the object is simply assigned to the class of its nearest neighbor. In binary (two class) classification problems, it is helpful to choose k to be an odd number as this avoids difficulties with tied votes. The same method can be used for regression, by simply assigning the property value for the object to be the average of the values of its k nearest neighbors. It can be useful to weight the contributions of the neighbors, so that the nearer neighbors contribute more to the average than the more distant ones. The neighbors are taken from a set of objects for which the correct classification (or, in the case of regression, the value of the property) is known. This can be thought of as the training set for the algorithm, though no explicit training step is required. In order to identify neighbors, the objects are represented by position vectors in a multidimensional feature space. It is usual to use the Euclidean distance, though other distance measures, such as the Manhattan distance could in principle be used instead. The k-nearest neighbor algorithm is sensitive to the local structure of the data.
The code has been tested with AT&T database achieving an excellent recognition rate of 99.20% (40 classes, 5 training images and 5 test images for each class, hence there are 200 training images and 200 test images in total randomly selected and no overlap exists between the training and test images).
Index Terms: Matlab, source, code, face recognition, face matching, face verification, dct, k-nearest neighbor algorithm, knn, discrete cosine transform.
 |
Figure 1. Example of k-NN classification |
||||||||||||||
|
A simple and effective source code for Face Recognition. |
|||||||||||||||
 |
Demo code (protected P-files) available for performance evaluation. Matlab Image Processing Toolbox is required. |
||||||||||||||
|
Release
|
Date
|
Major features
|
|||||||||||||
|
1.0
|
2007.10.20 |
|
|||||||||||||
|
We recommend to check the secure connection to PayPal, in order to avoid any fraud. |
|||||||||||||||
|
Face Recognition Based On Overlapping DCT – Click here for your donation. In order to obtain the source code you have to pay a little sum of money: 250 EUROS (less than 350 U.S. Dollars).
|
|||||||||||||||
|
Once you have done this, please email us luigi.rosa@tiscali.it
As soon as possible (in a few days) you will receive our new release of Face Recognition Based On Overlapping DCT. Alternatively, you can bestow using our banking coordinates:
|
|||||||||||||||
The authors have no relationship or partnership with The Mathworks. All the code provided is written in Matlab language (M-files and/or M-functions), with no dll or other protected parts of code (P-files or executables). The code was developed with Matlab Release 2006a. Matlab Image Processing Toolbox is required. The code provided has to be considered “as is” and it is without any kind of warranty. The authors deny any kind of warranty concerning the code as well as any kind of responsibility for problems and damages which may be caused by the use of the code itself including all parts of the source code.
Popularity: 3% [?]
K-Nearest Neigbors and Radius (Range) Search
Filed under: Algorithms, Computational Geometry, Mathematics
BruteSearchMex
When the dataset is small, when you have to run only a few number of search, or when the dimensions of points is large, the brute search method is still faster than kd-trees data structure. Computing the distances one by one take a minor time than building the tree.
Some of this problems have became less gravous since the introduction of GLTree (file id 22190) which allows for a very fast tree construction.
Despite this, very small dataset are still terrain for brute search algorithms.
I saw many k-neighbours utilities on FEX, but all them were m-coded. I think such a brute calculation is not an m-code job. So I developed my own Nearest Neighbour finder. It is nothing special it just computes all the distances and take the ones required from the input parameters, but of course the mex implementation make it even faster than vectorized m-code.
The following utilities are provided:
- Nearest neighbor
- K-Nearest neighbors
- Radius Search
They al supports N-dimensions and work on double, it is possible to choose if return the distances.
Here is a time comparison with a vectrized m-code:
Output:
Elapsed time is 0.962640 seconds.
Elapsed time is 18.813100 seconds.
For info, questions, suggestions, bugs: giaccariluigi@msn.com
Download Now
Popularity: 42% [?]
Translator
By N2H Follow Us
Video Posts
-
