Adaptive Denoising techniques for Medical Images in Wavelet Domain (original) (raw)
Medical Images generally have poor contrast and complex nature of noise, as the noise is generated due to various acquisition devices and algorithm used. So denoising of medical images is a particularly delicate and difficult task. Wavelets provide an orthonormal basis for multiresolution analysis and decorrelation of non-stationary time series and spatial process. The decorrelation property of wavelets makes them suitable to denoise the images. There are three basic approaches for denoising: based on computation of simple threshold, regularity detection in wavelet domain and bayesian estimation in wavelet domain. The present work compares these techniques in term of Mean Square Error and discusses the relative advantages and disadvantages for their application to medical images. Simple threshold based methods are simple to implement and they perform good results. Regularity detection based techniques are better for the boundary detection of object as well as denoising. Bayesian estimation based methods perform very well for images where some information about distribution of noise is known. They will perform poor in some cases where the unknown type of noise is present.
Loading Preview
Sorry, preview is currently unavailable. You can download the paper by clicking the button above.