Aimed at the problem that long-term coherent integration is troublesome for the range migration of high velocity moving targets, the Keystone transform is employed for compensating the range migration and is an efficient method for improving the performance of high velocity targets detection and ISAR imaging. However, the implementation of Keystone transform in real time is very difficult for the high calculation complexity. In this paper, a configurable parallel architecture for the real-time Keystone processing is proposed, which can achieve the resources and performance exchange, by increasing the parallel number to improving the performance. The result of simulation and verification on board show that our architecture for the Keystone processing is effective, comparing to the theoretical result, the maximal relative error of which is less than 10e-8, when used to accomplish the coherent integration of high velocity targets, and which need about 25μs to achieve single frequency Keystone transform with 2048 pulses in average, meeting the real-time processing requirement, when setting the parallel number is 1.