Multi-Rate Signal Processing with the Use of Filter Banks Composed of Parallel FIR Filters

Abstract

The paper presents a concept of a hardware implementation of 2-D finite impulse response (FIR) filter banks for the application in image processing and analysis. Banks composed of low-and high-pass FIR filters are basic components of multi-stage discrete wavelet transform (DWT). The applications of such solutions that are in the scope of our interests are vision systems used in automotive active safety functions (e.g in line departure warning). Basics of the DWT are broadly described in the literature. In our work we focus on solutions supporting hardware realization of filter banks for DWT. The proposed parallel and asynchronous circuits allow to achieve the processing time for a single pixel not exceeding 2 to 4 ns, depending on the size of the mask (data for TSMC 180 nm CMOS process).