|
|
Development of algorithms for digital real time image processing on a DSP Processor
Knapo, Peter ; Sajdl, Ondřej (oponent) ; Belgium, Jurgen Baert (MSc), KHBO (vedoucí práce)
Face recognition is a complex process that aims to recognize human faces in images or video sequences. Applications include surveillance and identification system, but face recognition is also invaluable in the research of computer vision and artificial intelligence. Face recognition systems are often based on either image analysis or neural networks. This work implements an algorithm based around the use of so-called eigenfaces. Eigenfaces are the result of a form of Principal Component Analysis (PCA), which extracts important facial features from the original image and is based on solving a linear matrix equation of the covariance matrix, eigenvalues and eigenvectors. A face that is to be recognized is thus projected onto the eigenspace; the results of that operation can be interpreted as the comparison of this face with an existing database of known faces. Before executing the actual recognition algorithm, faces need to be located inside the image and prepared (by doing normalization, lighting compensation and noise removal). Many algorithms exist, but this work uses a color based face detection algorithm, which is both fast and sufficient for this application. The face detection and recognition algorithms are implemented on a Blackfin ADSP-BF561 DSP processor from Analog Devices.
|
|
|
Development of algorithms for digital real time image processing on a DSP Processor
Knapo, Peter ; Sajdl, Ondřej (oponent) ; Belgium, Jurgen Baert (MSc), KHBO (vedoucí práce)
Face recognition is a complex process that aims to recognize human faces in images or video sequences. Applications include surveillance and identification system, but face recognition is also invaluable in the research of computer vision and artificial intelligence. Face recognition systems are often based on either image analysis or neural networks. This work implements an algorithm based around the use of so-called eigenfaces. Eigenfaces are the result of a form of Principal Component Analysis (PCA), which extracts important facial features from the original image and is based on solving a linear matrix equation of the covariance matrix, eigenvalues and eigenvectors. A face that is to be recognized is thus projected onto the eigenspace; the results of that operation can be interpreted as the comparison of this face with an existing database of known faces. Before executing the actual recognition algorithm, faces need to be located inside the image and prepared (by doing normalization, lighting compensation and noise removal). Many algorithms exist, but this work uses a color based face detection algorithm, which is both fast and sufficient for this application. The face detection and recognition algorithms are implemented on a Blackfin ADSP-BF561 DSP processor from Analog Devices.
|
host ::
RSS.