A Novel Thresholding Technique for Adaptive Noise Reduction using Neural Networks

The speckle corrupted image is a traditional problem in both biomedical and in synthetic aperture processing applications, including synthetic aperture radar (SAR). In a SAR image, speckle manifests itself in the form of a random pixel-to-pixel variation with statistical properties similar to those of thermal noise. Due to its granular appearance in an image, speckle noise makes it very difficult to visually and automatically interpret SAR data. Therefore, speckle filtering is a critical preprocessing step for many SAR image-processing tasks, such as segmentation and classification. Wavelet multiresolution analysis has the very useful property of space and scale localization, so it provides great promise for image feature detection at different scales. The recent wavelet thresholding based denoising methods proved promising, since they are capable of suppressing noise while maintaining the high frequency signal details. However, the local space-scale information of the image is not adaptively considered by standard wavelet thresholding methods. In standard wavelet thresholding based noise reduction methods, the threshold at certain scale is a constant for all wavelet coefficients at this scale. In this paper varies thresholding techniques have been studied for adaptive noise elimination and we presented a new type of thresholding neural network (TNN) structure for adaptive noise reduction, which combines the linear filtering and thresholding methods. This method produced better results than traditional methods.