Title :
Enhanced real time facial detection and replacement using GPGPU
Author :
Lancaster, Mark ; Green, Ron
Author_Institution :
Dept. of Electr. & Comput. Eng., Univ. of Canterbury, Christchurch, New Zealand
Abstract :
A method is presented for enhanced facial detection and replacement. This method builds on previous methods by addressing three key shortcomings: low performance, a tendency to misidentify walls as faces, and poor blending of the replacement face with the host body. General purpose graphics processing unit (GPGPU) technology is used to make this method feasible for use. Parameters for the algorithm are determined by experimentation. Measurements of the system show up to 3 times performance improvement from using the GPU. Blending of the host and replacement is smoother, though not seamless. Walls appear to no longer be misidentified as faces.
Keywords :
face recognition; graphics processing units; GPGPU technology; enhanced real time facial detection; facial replacement; general purpose graphics processing unit technology; Cameras; Face detection; Face recognition; Filtering algorithms; Graphics processing units; Image color analysis; Image resolution; Computer vision; face detection;
Conference_Titel :
Image and Vision Computing New Zealand (IVCNZ), 2013 28th International Conference of
Conference_Location :
Wellington
Print_ISBN :
978-1-4799-0882-0
DOI :
10.1109/IVCNZ.2013.6727029
