Acceleration of a locally tuned sine non linear video enhancement algorithm on GPGPU
Date of Award
2011
Degree Name
M.S. in Electrical Engineering
Department
Department of Electrical and Computer Engineering
Advisor/Chair
Advisor: Tarek Taha
Abstract
Computer Vision based applications support various domains such as medical, manufacturing, military intelligence and surveillance systems. These applications can be divided into: image acquisition, pre-processing, feature extraction, detection or segmentation, and high-level processing. However these tasks are time intensive due to the compute bound nature of the algorithm. In this thesis, an algorithm, based on an image dependent nonlinear function, the Locally Tuned Sine Nonlinearity (LTSN), is accelerated using NVIDIA's Computer Unified Device Architecture (CUDA) and the CPU. The main core of the algorithm is a nonlinear sine transfer function which is very flexible in enhancing the dark regions and compressing overexposed regions of an image. The video enhancement algorithm gave 21 frames per second compared to 9 frames per second for a 480p video. It is envisaged that the new technique would be useful for improving the visibility of scenes of night time driving and night security situations in real time.
Keywords
Night vision devices Research, Photography Exposure Calibration, Image processing Digital techniques
Rights Statement
Copyright © 2011, author
Recommended Citation
John, Julian Daniel, "Acceleration of a locally tuned sine non linear video enhancement algorithm on GPGPU" (2011). Graduate Theses and Dissertations. 398.
https://ecommons.udayton.edu/graduate_theses/398