Saturday, September 4, 2010

Video based Face Recognition using GPU

Face recognition has become a popular area of research in computer vision and one of the most successful applications of image analysis and understanding. Video-based face recognition systems consist of three modules:
1.A detection module
2. a tracking module
3.Arecognition module.
Given a frame of a video sequence, the detection module locates face candidates, while the tracking module finds the exact position of facial features in the current frame based on an estimate of face or feature locations in the previous frame. The recognition module identifies or verifies the face, integrating information from previous frames.
In the detection module, motion and/or skin color information can be used for segmenting the face from the background and locate candidate face regions. Face detection techniques similar to those applied for still images are then employed to find the exact location of faces in the current frame, thus initiating face and facial feature tracking. Face tracking techniques include head tracking, where the head is viewed as a rigid object performing translations and rotations, facial feature tracking, where facial features deformations due to facial expressions or speech are viewed as non rigid transformations limited by the head anatomy, and methods tracking head and features. Face and facial feature tracking is sometimes used to reconstruct the 3D shape of the face, which is subsequently used for enhancing face recognition.
The main problem of video-based face recognition is low quality of images in video sequences, while the unquestionable advantage is the abundance of information. This enables the selection of the frames that will be used for recognition and the reuse of recognition information obtained in precedent frames. Also, temporal continuity allows tracking of facial features, which can help in compensating pose or expression variations, while motion, gait and other features may enhance the performance of face recognition, Moreover, the simultaneous comprehensive exploitation of spatiotemporal cues results in increased tracking and identification accuracy. There are a number of algorithms for face detection, and recognition. Such as algorithms based on PCA(Principal Component Analysis) , LDA(Linear Discriminant Analysis), ANN(Artificial Neural Network), Viola and Jones face detection algorithm etc . Out of all these algorithm Viola and Jones face detection algorithm is the one best suited for the real time face recognition.
PROBLEM:
 Variations in lightening conditions, low quality videos, small size of acquired faces and variations in face expressions and pose are the main problems encountered by the video based face recognition system. Face detection algorithm proposed by Viola and Jones has made face detection practically feasible in real world applications such as digital cameras and video surveillance system. The Haar features used in face detection algorithm by Viola and Jones are very simple and effective for frontal face detection, but they are less ideal for faces at arbitrary poses.
SOLUTION:
Following are the solutions proposed for the better functioning and enhancement of video based face recognition System:
1.      Viola and Jones face detection algorithm can be modified to work well for the arbitrary poses by using the shoulder as another feature to detect the face, and exploiting the uniqueness of the earlobe for the detection. Further by using GPU we can modify/replace the Haar features to some complex features.
2.      For Face recognition we can take advantage of genetic algorithm to search the given face within the database and provide a quick output.
3.      We can further Use Dynamic thread scheduling on Multiple GPUs to enhance the working of Face recognition system for detecting and processing images with multiple faces.

Friday, August 27, 2010

Hi, frnds this is my first post .. so just hi , hello and intro.....
this post is dedicated to all the frustration I might be going thru during next 10-12 months becoz of my Masters project .
here i will be posting what i did nd what i m going to do blah, blah, blah,...
if u ppl have any ideas regarding this or ny other thing , do convey them to me ,
and 1 more thing, sometimes I might post something interesting findings of mine from net surfing .. so bear them with me....
lastly welcome to re-searcher@work....
hope this blog will be entertaining as well as educational for all of us ....:) :)