In this work we revisit TV filter and propose an improved version that is tailored to diagnostic CT purposes. We revise TV cost function, which results in symmetric gradient function that leads to more natural noise texture. We apply a multi-scale approach to resolve noise grain issue in CT images. We examine noise texture, granularity, and loss of low contrast in the test images. We also discuss potential acceleration by Nesterov and Conjugate Gradient methods.
CT image quality is affected by various artifacts including noise. Among these artifacts of different causes, noisy data
due to photon starvation should be contained in early processing stage to better mitigate other artifacts as they can cause
severe streaks and noise in reconstructed CT image. For low dose imaging, it is critical to use effective processing
method to handle the photon starved data in order to obtain required image quality with desired resolution, texture, low
contrast detectability. In this paper, two promising projection domain noise reduction methods are proposed. They are
derived from (1) the noise model that connects the noise behaviors in count and attenuation; (2) predicted noise
reduction from a finite impulse response (FIR) filter; (3) two pre-determined noise reduction requirements (noise
equalization and electronic noise suppression). Both methods showed significant streaks and noise reduction in tested
cases while reasonably maintaining the resolution of the images.
Sub-pixel compounding is a technique that synthesizes the information of an image sequence to form a betterresolved
and speckle reduced image. To avoid extra data acquisition time and patient exposure, reuse of the existing data
is highly desired. In elasticity imaging, a set of images with slight changes due to deformation is produced, which
provides an ideal input for the sub-pixel compounding process. In this paper, a brief review of the resolution
enhancement techniques in ultrasound imaging will be provided, and then, a diffusion-regularized, least square approach
is presented for sub-pixel compounding image reconstruction. Based on the results, we suggest that (1) B-mode images
from elastic imaging are suitable data for sub-pixel compounding and a speckle noise reduced higher-resolution image is
a co-product of elasticity imaging; (2) for breast diagnosis, resolution improvement is of strong interest since better
depiction of the interior and exterior structures of a tumor provides important detection and diagnostic information; (3) a
similar approach could be extended to elasticity imaging with other modalities.
The intima-media thickness (IMT) of the carotid artery is an important biomarker for the clinical prognosis and diagnosis of atherosclerosis and stroke. This paper presents a new approach, pixel compounding, to enhance the resolution of the intima-media vascular layers in ultrasound B-scan images and provide increased image resolution for a more precise measurement. First, homomorphic transformation is used to estimate the lumped point spread function (PSF) of the images, then, the images are deblurred with the estimated PSF, and finally, a non-homogeneous anisotropic diffusion algorithm is used to further enhance the resolution of the image. The homogeneous part of the algorithm is used to suppress speckle while enhancing the coherent structures, specifically the edges; the non-homogeneous part (likelihood estimator) progressively adds the details from succeeding frames in the image sequence for an optimal and sub-pixel resolved image. Phantom studies have shown 300% improvement on Peak Distance Standard Deviation and nearly 100% improvement on Average Half Peak Width, indicating significant resolution enhancement.
One factor used to tell if a tumor is benign or malignant is by examining the underlying structures of the tumor. In ultrasound images, these underlying structures are often buried in speckle noise. The proposed technique utilizes a median-anisotropic diffusion interactive algorithm to “dissolve” speckle noise and enhance images. This technique has two important differences from conventional anisotropic diffusion techniques. First, there is a two resolution level process, which converts the speckle to quasi-impulsive noise, making it more easily removable. Second, a median related reaction term is applied to the anisotropic diffusion equation, thus making it well suited for removing impulse type noise. In addition, the reaction term will make speckle reduction and image enhancement more adaptive to the local features in the image. The experimental data showed the proposed technique improved the legibility of the breast tumor images as important structures were emphasized.