SOPHIA VILLALBA - MATLAB Image Filtering

MATLAB Image Filtering

Skills: MATLAB

Purpose

Biomedical imaging devices employ various techniques to capture the intricacies of the human body. X-rays and CT scans require electromagnetic radiation, ultrasounds use sound waves, MRIs use the magnetic field and radio waves, and nuclear imaging use radioactive tracers. All of these devices process raw data and then perform various calculations to present it in the visual form of an image. Each method has its unique downsides when it comes to image quality and its own filtering for noise and fuzziness in the image. For this project, I studied different filtering methods and determined which are more useful for each imaging device.

Original Image

For the purpose of this project, the original image that will be recreated is simply three squares in a matrix of zeros. This will demonstrate the impact of noise around actual artifacts and how it affects the image. The gradient will visualize how similar the filtered image is to the original image.

Low-Pass vs High-Pass Filtering

Low-pass filtering allows low frequencies through and blocks high frequencies. In imaging, this translates to smoothing noise and harsh edges, creating a blurry version of the original image. As shown in the pictures generated, as the filter size increased, the image more closely resembled the original.

High-pass filtering creates a cutoff to allow for high frequencies and limits low frequencies. This filtering results in sharper edges and less uniform areas, good for outlining the objects in the original image. The pictures demonstrating different filter sizes have sharper lines as the filter size increases, with the 0.5 filter size showing only the hight point of contrast at the corners of each of the squares.

Low-pass filtering is more useful in ultrasound imaging where the fine line details of the outline are not as important as reducing the background noise of the soft tissue imaging of organs and blood vessels. High-pass filtering is useful for when precise boundaries of - organs and tissues are necessary, like in nuclear medicine imaging, especially SPECT imaging where exact mapping is required.

Butterworth High-Pass Filtering

There are different filters that can be applied to these images to further refine and improve image quality. The Butterworth filter minimizes ripples and distortion within a specified frequency range. Comparing this filter to the default high-pass filtered image of object filter 0.1, the noise is significantly reduced, and the lines are much sharper. The Butterworth High-Pass filter results in less noise around the edges of the objects because it removes the lower frequency components as opposed to the high-pass filter. The ideal high-pass filter features aliasing around the corners, which contributes to the additional noise compared to the Butterworth filter.