Dual Stage Bayesian Network with Dual-Tree Complex Wavelet Transformation for Image Denoising
Image de-noising always plays a vital role in various engineering bids. Moreover, in image processing technology, image de-noising statistics is persisted as a substantial dispute. Over the past decades, certain de-noising methods have been exposed incredible accomplishments. This paper intends to develop a de-noising algorithm for multimodal and heterogeneous images, while the conventional de-noising algorithms handle a specific image type. The filtered information is reversed to spatial domain to recover the de-noised image. Dual tree Complex Wavelet Transform (DT-CWT) is exploited for image transformation for which the wavelet coefficients are estimated using Bayesian Regularization (BR). To ensure the de-noising performance for heterogeneous images, the statistical and wavelet features are extracted. Subsequently, the image characteristics are combined with noise spectrum to develop BR model, which estimates the wavelet coefficients for effective de-noising. Hence, the proposed de-noising algorithm exploits two stages of BR. The first stage predicts the image type, whereas the second stage estimates appropriate wavelet coefficients to DT-CWT for de-noising. The performance of the proposed model is analysed in terms of Peak Signal to Noise Ratio (PSNR), Second derivative Measure of Enhancement (SDME), Structural Similarity (SSIM), Mean Squared Error (MSE), Root Mean Square Error (RMSE), Mean Absolute Error (MAE), Pearson Coefficient (PC), and Symmetric Mean Absolute Percentage Error (SMAPE) respectively. The proposed model is compared to the conventional models, and the significance of the developed model is clearly described.