Dynamic Fuzzy Contol for Facial Image Denoising from Rough Images
Young Woon Woo
Doo Heon Song
Kwang Baek Kim
Abstract
Image denoising aims to restore the latent clean image from its noise-corrupted version. In this paper, we propose a method to restore facial image from damaged rough photograph images using Hopfield Neural network for fast and robust image denoising. From our previous experience of information loss during the binarization process, we avoid such binarization in the proposed method. Instead, we design a dynamically controlled fuzzy stretching to enhance the intensity contrast and to handle rough intensity distribution. We also modify Hopfield network to use Max-Min fuzzy operation to learn a continuous value between 0 and 1 to improve the robustness. The experiment verifies the efficacy of the proposed method