A New Image Restoration Algorithm Based on Mathematical Morphology and Wavelet Neural Network
Yan Shen, Ming Liu, Jinbao Wang, Wenlu Zhou · 2012
Wavelet neural network (WNN) is introduced into the field of image restoration due to the excellent local feature and adaptive ability.The procedure of restoration can be looked as an approximating procedure from noisy image to original image.The better WNN has approximation performance, the better restoration performance.With the help of wavelet neural network and mathematical morphology, a new image restoration algorithm is proposed.It can effectively maintain the image edges and details.In order to overcome the shortcomings of the filter with fixed structure, amoeba structure element is presented based on mathematical morphology.Image data is extracted by amoeba structure element, and input into the wavelet neural network .The WWN is trained by BP algorithm in batch mode training, which adjusts the wavelet coefficient and network weights adaptively.The experimental results showed that the approach proposed can preserve fine details and excellent fidelity and is better than general denoising methods.