An Efficient High-definition Real-Time Hardware Design for Bilateral Filtering

Bilateral filtering is a simple nonlinear video denoising method. As an effective edge preserving or denoising method, bilateral filtering requires less amount of data as input and involves simpler operation than spatiotemporal filtering. In addition, it can preserve spatial details well and realize satisfactory visual quality. However, bilateral filtering involves complex processing and is data-dependent, which is not conducive to hardware implementation. To address these problems, in this study, we proposed a high-definition real-time hardware design method for bilateral filters, including parallelism, look-up tables, and other design methods. By using the proposed design system, the data dependence of bilateral filtering can be removed, and the throughput can be improved. Furthermore, we developed an efficient field programmable gate array implementation of an edge-preserving fast bilateral filter on a hardware-software co-design environment. Results demonstrated that the goal of 1080p @ 30 fps was reached, and the number of cycles was reduced to 96, thus, greatly improving the filtering process. Moreover, video filtering speed plays a crucial role in video streaming.