Reconfigurable Discrete Hilbert Transform Algorithm For VLSI

The rapid development of modern medicine, came into being multi-slice spiral CT, spiral cone-beam scanning methods are the reconstruction speed advantage of high spatial resolution. But the bottleneck is also very clear: in spiral cone-beam CT three-dimensional image reconstruction, the computer and the receiver between a large number of different points of data through the discrete Hilbert transform to the transfer, so that accurate image reconstruction. The use of the past, discrete Hilbert Transform method, using a fixed point of the VLSI hardware implementation and can not deal with different points of data. Systolic Array-based algorithms can be configured, you can solve this problem.This paper is divided into two parts, the first two chapters of the first part, described the current status of the development and use of CT configurable systolic array design of the necessity of Hilbert filter is introduced Katsevich algorithm is three-dimensional spiral cone-beam CT reconstruction principles and DHT The definition and a number of design approach, focusing on the systolic array design approach.After three chapters of the second part, is also the focus of this article. In the third chapter, through the derivation of systolic array design demonstrates the feasibility of the method can be configured. The fourth chapter, for the third chapter of the program triggered by parameters of the problem, made a further design study made a more comprehensive reasoning and sorting have been calculated based on the boundary parameters can be configured to the calculation method, and various boundary parameters of the calculation formula. And finally be the method and the original method in all cases carried out a series of theoretical comparison, confirmed the complexity of the algorithm, the new method does have its advantages: The new method is always better than the space complexity the original method. In addition the unit will be submitted separately PE calculation algorithms, as long as the N value reached 32 points or more in size, the time complexity of the new method is also superior to the original method.Finally, the results of this paper has conducted a summary of the existing problems were analyzed, and further study are given suggestions and ideas.