The convolution operation based two dimensional (2D) filters for biomedical image preprocessing are computationally intensive and consume significant hardware resources on Field Programmable Gate Array (FPGA). The applicability of these filters for different clinical scenario demands multiple types and sizes of 2D filters which proportionally increases the hardware consumption. In this work a fully reconfigurable 2D filter is proposed featuring flexible functionality to eliminate the requirement of implementing multiple filters separately for each clinical-scenario. It is capable to configure its hardware at runtime for performing user defined filtering operation on input image by choosing appropriate coefficients from pre-stored filter coefficients lookup table. The hardware requirement of design is reduced by exploiting intra structural similarities of 2D filters which eliminate the redundant filter-coefficients storage and its associated hardware for selection circuit. Our proposed design approach significantly reduces the hardware implementation cost of biomedical image pre-processing as compared to the conventional approaches (i.e. multiple hardware for implementing multiple filters separately, each with fixed functionality). This reduction in hardware cost consequently reduces its power consumption and enhances design performance hence it is capable to support high frame rates (1024x1024@201fps) at a very low total power consumption of 170 m Watt. All these attributes makes it very suitable for resource constraint applications such as Point of Care devices. The design is fast prototyped and validated using Xilinx System Generator on smallest Xilinx Atrix-7 (XC7A35T) device.
2D Filters; Biomedical Image Preprocessing; Reconfigurable Hardware; Xilinx System Generator; Resource Constraint Applications.