[Example] Automatic parallelization for image processing programs
Introducing a case study on the acceleration of image processing programs using the OSCARTech compiler.
The "OSCARTech Compiler" supports automatic parallelization of a wide range of software programs. We will introduce a case of automatic parallelization applied to a fundamental image processing algorithm called "Optical Flow," which detects the magnitude of movement (motion vectors) of objects within an image. In this case, we attempted parallelization using block matching (a method where blocks are cut out from a still image at a certain moment, and the vicinity of the same coordinates in the next moment is searched to select and determine the most similar block). 【Case Study】 ■ Optical Flow: Detects the direction and magnitude (motion vectors) of movement of objects within an image. ■ Verification Environment: Raspberry Pi 2 (Cortex-A7, 4 Core, 900Mhz, OS: Raspbian OS) ■ Time taken to generate motion vectors: - 1 Core: 936ms / 1.0x - 2 Cores: 480ms / 1.9x - 3 Cores: 340ms / 2.7x *For more details, please refer to the PDF document or feel free to contact us.
- Company:オスカーテクノロジー
- Price:Other