An Innovative Wavelet Neural-network Algorithm For Recovering Damaged Blocks In High-resolution Digital Images

Journal of Applied Science and Engineering

Alaa K. Al-azzawi¹ 

[1] Department of Electronics & Communications Engineering, Technical Engineering College-Baghdad, Middle Technical University, Ministry of Higher Education Scientific Research, Baghdad-Iraq;

关键词: wavelet network;    wavelet decomposition;    discrete wavelet transform (dwt);    artificial neural network (ann);    blurring;    artifacts;   

DOI  :  10.6180/jase.202302_26(2).0001

来源: DOAJ

【 摘 要 】

Wavelet transformations with neural networks can be used to classify and identify the most important problems that may occur when analyzing high-resolution digital images in a distinctive style. In this paper, the discrete wavelet transformations (DWT) were adopted, after using a 3-levels of Haar decomposition to decompose the damaged images. The lost coefficients in the high frequency sub-bands of the 3-haar levels were guessed by using the vertical and horizontal interpolation process between the lost and their adjacent pixels. Evaluation results for these coefficients were more accurate after calculating the mean square errors at the top and bottom of the missing pixels, respectively. Further, the estimated decomposition matrices were directly connected with a trained artificial neural network (ANN) in order to increase the accuracy of the results and obtain high quality images. The artificial neural network architecture was trained in an efficient configuration and represented by a fast forward multi-layer perceptron using resilient back-propagation with the intention of reducing error ranges (i.e., blurring and artifacts). Experimental results were convincing and very close to the desired values.

【 授权许可】

Unknown