Traffic sign image denoising with energy-based adaptive finite ridgelet transform

Denoising of traffic sign images is an important pre-processing step in intelligent transportation system whose performance greatly affects subsequent manipulations. Based on characteristic analysis of traffic sign images that linear singularities is more informative, this paper introduced the finite ridgelet transform (FRIT) for traffic sign image denoising. An energy based adaptive finite ridgelet transform scheme (EFRIT) is proposed for better presentation of linear singularities. Abundant experimental results carried out on different types of traffic sign images at varying noise levels demonstrate the effectiveness of the proposed algorithm in both terms of PSNR and visual quality.