MEDICAL IMAGE ENHANCEMENT USING TRANSFORM BASED METHODS

Abstract

Nowadays, clinicians are allowed to noninvasively evaluate physiological processes within the human body by means of various medical imaging modalities, such as computed tomography, positron emission tomography, functional magnetic resonance imaging, and ultrasonography. Among them, ultrasound imaging is definitely the one that offers the best price-to-performance ratio receiving thus, an added attention. However, one major issue when using this imaging modality is the inherent presence of speckle noise. Its occurrence is often undesirable, since it effects the tasks of human interpretation and diagnosis. On the other hand, its texture carries important information about the tissue being imaged. Speckle filtering is therefore a critical pre-processing step in medical ultrasound imagery, provided the diagnostic features of interest for diagnosis are not lost. The quality of the image can be improved by using speckle reduction (filtering) techniques. Adaptive filtering techniques can be used for this purpose but the main drawbacks in adaptive techniques are "edge preserving" and "feature preserving" of the image. Since these filters are sensitive to the size and shape of the filter window, leading to either over-smoothing of the image edges or incomplete reduction of speckle in the ultrasound image. In order to overcome these limitations of adaptive filtering techniques, recently, the wavelet and curvelet transforms has been proposed as a useful processing tool for signal recovery. However, while filtering an image using wavelet or curvelet transforms separately, wavelet transform provides unique advantages in dealing with singular point, as well as some disadvantages in dealing with the borders and the line features of the image. Curvelet transform can be used for the finest approximation of the singular curve in the image. In this thesis, we combine the wavelet transform and curvelet transform to deal with the noisy ultrasound images. The results show that the proposed method is better than only using cavelet transform or curvelet transform to deal with noisy image, and this method achieves excellent visual effects and higher SNR with MSE drastically reduced, compared to wavelet and curvelet filters.