eduzhai > Applied Sciences > Engineering >

Enhancement of a CNN-Based Denoiser Based on Spatial and Spectral Analysis

  • Save

... pages left unread,continue reading

Document pages: 5 pages

Abstract: Convolutional neural network (CNN)-based image denoising methods have beenwidely studied recently, because of their high-speed processing capability andgood visual quality. However, most of the existing CNN-based denoisers learnthe image prior from the spatial domain, and suffer from the problem ofspatially variant noise, which limits their performance in real-world imagedenoising tasks. In this paper, we propose a discrete wavelet denoising CNN(WDnCNN), which restores images corrupted by various noise with a single model.Since most of the content or energy of natural images resides in thelow-frequency spectrum, their transformed coefficients in the frequency domainare highly imbalanced. To address this issue, we present a band normalizationmodule (BNM) to normalize the coefficients from different parts of thefrequency spectrum. Moreover, we employ a band discriminative training (BDT)criterion to enhance the model regression. We evaluate the proposed WDnCNN, andcompare it with other state-of-the-art denoisers. Experimental results showthat WDnCNN achieves promising performance in both synthetic and real noisereduction, making it a potential solution to many practical image denoisingapplications.

Please select stars to rate!

         

0 comments Sign in to leave a comment.

    Data loading, please wait...
×