1. A NOVEL METHOD FOR COLOR FACE RECOGNITION USING KNN CLASSIFIER
J.Prabin jose P.Poornima K.Manoj Kumar
Ravi Teja Kotaprolu

2. Contents: Introduction Proposed method Methods Result and Discussions
Color space models
Principal component analysis
Nearest neighbor classifier
Result and Discussions
Conclusion and Future work

3. Introduction
Digital image is composed of a finite number of elements, each of which has a particular location value. These elements are referred as picture elements, image elements, pixels.
The digital image processing is the use of computer to perform image processing on digital images.
Digital image processing has many advantages over analog image processing. It allows a much wider range of algorithms to be applied to the input data and can avoid problems. Such as build-up of noise and signal distortion.

4. Contd..
Recently face recognition(FR) has become an important area of research in computer vision, neuroscience and psychology .
The effectiveness of color information plays an important role when face images are taken under strong variations in illuminations as well as with low spatial resolutions for improving FR performance.
In general, the three components of a color image can be defined in many different ways leading to a wide variety of color spaces. The RGB is a fundamental and most widely used color space. Other color spaces are obtained by transformations of RGB.

5. PROPOSED METHOD The flowchart of the proposed method :

6. Contd..
In this method the given image is transformed into various color space models like RGB, YCbCr, YIQ, HSV.
Then various color components available in the color space model are separated.
Then the Eigen value and vector features are extracted from the various color component images using PCA(Principal Component Analysis)
Finally the extracted features are used to classify test image by using nearest neighbor classifier.

7. METHODS: COLOR SPACE MODELS:
A color space model is a mathematical representation of a set of color.
The color space models:
RGB
YCbCr
YIQ
HSV

8. RGB:
YCbCr:

9. HSV: YIQ:
original
Luma (y) ->
Chroma (I) ->
Chroma (Q) ->

10. Principal Component Analysis(PCA)
Principal component analysis is a mathematical procedure that uses to convert a set of observations of possibly correlated variables into set of values of uncorrelated variables called principal components.
PCA is appropriate when you have obtained measures on a number of observed variables and wish to develop a small number of artificial variables.

11. Contd..
PCA is a quantitatively rigorous method for achieving the simplification of dataset.
It is because, in data set with many variables often move together. one reason is that more than one variable may be measuring the same driving principle governing the behavior of the system.
The method generates a new set of variables called principal components. Each principal component is a linear combination of the original variables.
Principle components are orthogonal to each other so there is no redundant information.

12. Steps that involved in PCA process:
Step 1 : Get the data. Step 2 : Subtract the mean. Step 3 : calculate the covariance matrix. Step 4 : calculate the eigen vectors and eigen values of covariance matrix. Step 5 : calculate the summation value of eigen vector and eigen values and form a feature vector.

13. Nearest Neighbour classifier
In pattern recognition , the k-nearest neighbour algorithm (k-NN) is a widely used classifier for classifying objects based on closest training examples in the feature space. The kNN algorithm is the simplest classifier of all machine learning algorithms. In this classifier image is classified by a majority votes of its neighbours. In kNN classifier the Euclidean distance between the testing image feature and each training image feature is determined to form distance matrix.

14. Contd..
The summation value of distance matrix is estimated and sorted in increasing order. The first k elements are selected and majority class value is determined for classifying the image accurately.

15. RESULTS AND DISCUSSIONS
The proposed method is implemented in matlab. In this method 70 different persons face images are used to test the performance of the system .Each person database contains 15 different facial expressions. Out of 15 face images 10 face images are considered as training feature. All the input images are converted into various color space models. The input images and various color components of the face image are shown :

17. Various color space model of the input image:

18. The eigen values and eigen vectors are extracted from each image.
The summation value of eigen vector and eigen value are considered as features of each color component image.
The kNN classifier is used to classify the different face images. In this paper 70 label values are created for 70 persons. The Euclidean distance between the testing image feature and the training image feature and a distance matrix is created.
In the distance matrix first k values are considered as and majority label of the k value is considered as the correct label of the given testing image.

19. The performance of given system is measured in terms of accuracy The accuracy is given by : Accuracy = 𝑐𝑜𝑟𝑟𝑒𝑐𝑡𝑙𝑦 𝑑𝑒𝑡𝑒𝑐𝑡𝑒𝑑 𝑓𝑎𝑐𝑒 𝑖𝑚𝑎𝑔𝑒𝑠 𝑡𝑜𝑡𝑎𝑙 𝑛𝑢𝑚𝑏𝑒𝑟 𝑜𝑓 𝑓𝑎𝑐𝑒 𝑖𝑚𝑎𝑔𝑒𝑠 In the classification problem 700 image feature value are considered as training feature and 350 image features are used for testing the classification.

20. Out of 700 training images 654 images are correctly classified and out 350 testing images 317 images are correctly classified without error. So, the overall accuracy obtained by this method is 92.47%.

21. CONCLUSION AND FUTURE WORK
In this paper , a novel and effective color FR method is proposed.
The image is transformed into various color space models. Then applying the PCA technique to every color space component and the features are extracted.
The features are used by kNN classifier for recognizing the input face image. The accuracy obtained in this method is much better than other results available in literature.
In future the optimal set of color features is found out from the color space for increasing the accuracy.

22. THANK YOU