Real-Time Processing of Ultrasound Images with Speckle Reducing Anisotropic Diffusion

Real-time image processing is desirable for many medical imaging applications. However, many sophisticated image processing algorithms are so computationally intensive that general-purpose CPUs cannot satisfy real-time requirements. In this work, we present a design flow graph approach and several design considerations for implementing complex image processing algorithms in hardware. As a case study, an FPGA implementation of a novel ultrasound image enhancement algorithm - speckle reducing anisotropic diffusion (SRAD) - is discussed. A considerable reduction of hardware complexity is achieved through the co-optimization of both the algorithm and its hardware implementation. Experimental results reveal a 30X performance improvement over the original MATLAB implementation and a 20X performance improvement over the C implementation, with a relative intensity error of less than 2%.