FETAL ULTRASOUND IMAGE DENOISING USING CURVELET TRANSFORM

The random speckle noise in the acquired fetal ultrasound images is caused by the interference of reflected ultrasound wave fronts. The presence of speckle noise will degrade the quality of the image and even hide image details, which in turn affect the process of image segmentation, feature extraction and recognition and most importantly disease diagnosis. The standardization of measurements from the fetal ultrasound images will help the physicians to make correct diagnosis. The accuracy of diagnosis is possible only when the image is noise free. Hence it is very much important to perform filtering of the speckle noise. It is proposed that curvelet transform serves as a better edge preserving filter compared to other speckle reducing anisotropic diffusion filters. Curvelet transform is designed to handle images which involve curves using only a less number of coefficients. Hence a multiscale representation called curvelet transform is applied to enhance the visual quality of the ultrasound images. The experimented results indicate that the proposed curvelet denoising suppresses the noise effectively both in quantitative and visual means by producing high PSNR.