Please use this identifier to cite or link to this item:
|Title:||Automatic photograph recognition||Authors:||Komathi Maturaveeran||Keywords:||DRNTU::Engineering::Electrical and electronic engineering||Issue Date:||2016||Abstract:||This report describes a novel algorithm for the recognition of a human face from the photograph of the person by detecting the location of the face and subsequently extracting its outline and essential features to transform into a sketch. Therefore, the algorithm has two main parts - face detection and face transformation. The former locates a face in the photograph by applying skin color segmentation combined with AdaBoost learning in an automatic manner. First, skin color model in the YCbCr color space is built to segment the non-skin color pixels from the image. Next, mathematical morphological operators are used to remove noisy regions and fill holes in the skin color region in order to extract the candidate human face region. Meanwhile, AdaBoost learning is used to classify images into faces and non faces. It achieves this by learning effective features from a large feature set; constructing weak classifiers, each of which is based on one of the selected features; and boosting these weak classifiers to construct a strong classifier. The weak classifiers are based on simple scalar Haar wavelet-like features. The strong classifier then classifies and detects the face in the desired photograph. The latter transforms the face into a sketch by extracting outlines of the face and its features in a semi-automatic manner whereby human intelligence is incorporated. For example, for the eyes and nose features, the user has to specify certain points of interest and some combination of thresholding is used with maximum gradient density techniques. For the lips, however, this explicit declaration of feature points is avoided. The user interaction process reduces to only specifying the region in which the feature is enclosed and the lips' orientation. The method of lips extraction is an almost automatic one and proves to be very successful.||URI:||http://hdl.handle.net/10356/68998||Rights:||Nanyang Technological University||Fulltext Permission:||restricted||Fulltext Availability:||With Fulltext|
|Appears in Collections:||EEE Student Reports (FYP/IA/PA/PI)|
Updated on May 8, 2021
Updated on May 8, 2021
Items in DR-NTU are protected by copyright, with all rights reserved, unless otherwise indicated.