The design and implementation of a reconfigurable board for rapid prototyping of a class of image processing algorithms

Previously, rapid prototyping approaches have addressed the needs of rapidly building specific hardware. We are taking a different approach to this problem and are viewing it as rapid prototyping of algorithms within a domain. This approach is useful when hardware acceleration is required, but the design entity is the image transformation algorithm rather than the hardware to perform the transformation. To experiment with this methodology we have designed and built boards which support a class of image processing algorithms that are dependent on streaming multiply-accumulates and table lookups.