Nonlinear wavelet filter for intracoronary ultrasound images

The development of a noise reduction filter for intracoronary ultrasound (ICUS) images is very important for the accurate detection of both lumen-intima (LI) and media-adventitia (MA) boundaries. In this study, the authors apply a nonlinear wavelet soft thresholding algorithm to ICUS images in order to enhance the detection of these boundaries. Three significant benefits can be achieved from this filtering algorithm: (1) the reconstructed images from thresholded wavelet coefficients is noise-free in the sense that no spurious oscillations are introduced; (2) boundary information in the image is preserved; (3) the wavelet transform procedure, which carries the authors' low-pass filtering and decimation, forces the image statistics from Rayleigh toward Gaussian. To compare the performance of the soft-thresholding algorithm as a preprocessing filter with other commonly used filters including the median filter, a local statistic filter and a Gaussian smoothing filter, the authors applied a standard boundary detection algorithm to a common set of images after preprocessing with each of the filters.