An infrastructure for analysing the performance of computational imaging algorithms on modern hardware

点击翻译按钮获取中文摘要

Numerous applications, such as those in medical imaging, film production, automotive and robotics, use imaging sensors (IS) to convert light to signals appropriate for further processing and storage by digital devices. IS output is imperfect and requires significant processing in the digital domain to yield acceptable results. For example, lens imperfections result in distorted output, while sensor sensitivity may yield output that is underexposed or overexposed at places and thus missing crucial information. Computational Imaging (CI) is the processing in the digital domain of IS output can compensate for these limitations. However, further advances in CI sophistication are severely limited by hardware performance. Specifically, at the heart of every computing device, there are one or more processing cores that manipulate data values and move them among storage devices or devices that interact with the physical world such as displays and touch screens. Existing general purpose processing cores are not keeping pace with the computational power demands of CI applications. For several decades, core performance, that is the rate at which data values could be manipulated or moved around, has doubled every two years. This has enabled advances in application sophistication and breadth. Application developers could rely on this trend of performance increase and plan ahead innovations to