An efficient hardware accelerated design for image denoising using Extended Trilateral filter

Abstract

Trilateral filtering presents an edge preserving smoothing filter. The predecessor of Trilateral filtering, the bilateral filter is a non-linear filtering technique that can reduce noise from an image while preserving the strong and sharp edges, but it cannot provide desired result when the edges have valley or ridge like features. The Trilateral filter is extended to be a gradient-preserving filter, including the local image gradient (signal plane) into the filtering process. This filter has the added benefit that it requires only one user-set parameter (neighborhood size used for bilateral gradient smoothing), and the rest are self-tuning to the image. In this paper, we introduce an extended version of Trilateral filtering where domain and range filtering are applied on the image separately followed by applying the original Trilateral filter on the image that is produced by combination of domain and range filtered outputs. This extended approach gives better noise reduction than the original Trilateral approach and improves the image quality. We also provide an efficient Field Programmable Gate Array (FPGA) based implementation of the proposed Extended Trilateral filter(ETF) on a hardware software co-simulation environment to validate the design.