Opencl based image processing library for embedded gpgpus and a sample face detection application

Since graphic processing units (GPUs) can be used for general purpose calculations, its spectrum of use increased. Not only GPUs can be used rendering, moving and simulating 3 dimensional graphics, but also can be used where data parallelism applies. GPUs include hundreds of cores, which enables them to process hundreds of threads simultaneously that standart processors can not achieve. Data parallel applications include image processing, linear algebra applications, fluid dynamics, bioinformatics and such. GPUs are used in these fields in order to accelerate common algorithms and applications. With emerging and downscaling new technologies, these GPUs can be used in mobile phone, tablets and various embedded systems. In this work, an image processing library has been developed and optimized for embedded GPGPU platforms. The library have been compared to the standardized library, OpenCV. An application of "Human Face Detection" which is widely used and its performance results are also the content of this work. Key Words: GPGPU, Embedded Software, Image Processing, Performance Optimization, Parallel Processing