US-Netherlands Cooperative Research: Fast Methods for ImageSequence Identification and Restoration

该奖项支持伦斯勒理工学院的Howard Kaufman教授和John W. Woods教授与荷兰代尔夫特理工大学电气工程系的Jan Biemond教授在计算机图像处理研究方面的合作。他们合作的目的是评估各种图像序列恢复技术在不同并行计算机体系结构上的有效性和效率。特别是，Biemond教授将使用基于迭代的程序和AP500阵列处理器，而Kaufman教授和Woods教授将使用基于递归的程序和sequential Balance 8000多处理器系统。空间和时间图像动态的识别将被两位研究者考虑。合作研究的另一个重要方面将是算法设计。重要的是，要使用良好的过程横截面，让所有处理器都有机会在自己的域中执行。同时，必须仔细分析实现的方法，以确保对体系结构能力的测量是公平的。畸变图像序列的快速恢复是一个日益重要的领域，需要有效的技术。由于最近硬件的进步，例如DSP芯片和并行架构，实时图像增强变得越来越可行。图像恢复通常是更密集的计算和更难以实现在高速。这种合作的目标是雄心勃勃的，但是对不同方法进行有意义的比较的预期结果将是非常值得的。

This award supports Professors Howard Kaufman and John W. Woods of Rensselaer Polytechnic Institute to collaborate in research in computer image processing with Professor Jan Biemond of the Department of Electrical Engineering of the Delft University of Technology, The Netherlands. The objective of their collaboration is to evaluate various image sequence restoration techniques in terms of their effectiveness and efficiency on different parallel computer architectures. In particular, Professor Biemond will use iterative based procedures and an AP500 array processor, while Professors Kaufman and Woods will use recursive based procedures and a Sequent Balance 8000 multiprocessor system. Identification of the spatial and temporal image dynamics will be considered by both investigators. Another important aspect of the cooperative research will be in algorithm design. It is important that a good cross-section of procedures be used to give all processors the chance to perform in their own domains. At the same time, the methods of implementation have to be closely analyzed to make sure that the measurements on the architecture's power are fair. The area of fast restoration of distorted image sequences is one of growing importance, for which effective techniques are needed. Real-time image enhancement is becoming increasingly more feasible due to recent hardware advances, such as DSP chips and parallel architectures. Image restoration is usually more computationally intensive and is more difficult to achieve at high speeds. The goals of this collaboration are ambitious, but the expected result of meaningful comparisons of different methodologies will be very worthwhile.