Image processing algorithms for medical applications : B

• Main Author: Muhammad Shamir Kamaluddin
• Other Authors: Mohammed Yakoob Siyal
• Format: Final Year Project
• Language: English
• Published: 2019
• Subjects: DRNTU::Engineering::Electrical and electronic engineering; DRNTU::Engineering::Computer science and engineering::Computing methodologies::Image processing and computer vision
• Online Access: http://hdl.handle.net/10356/78177
• Institution: Nanyang Technological University

In order to reduce the amount of noise in an image, image processing algorithms need to be carried out. This is an important procedure as image processing may result in noise corruptions either during transmission, acquisition or conversion. Such noise corruptions diminish the images’ visual quality and could also cause inaccuracy in the details of the image. Hence, there is a need for image processing algorithms, which restores an image such that the image’s visual quality is improved, producing a more precise details of the subject. In the field of medicine, in particular, having an accurate depiction and clear medical image is extremely crucial as doctors rely on medical images to analyze a patient’s internal structure and to diagnose accordingly. Should there be a lack of clarity in the image, it could potentially lead to inaccurate analysis and improper diagnosis. In this thesis, noise removal will be done using eight filters: Mean, Median, Gaussian, Wiener, Guided, Non-local mean, Diffuse and Bilateral filter. The efficacy of each filter will be analyzed using Signal to Noise Ratio (SNR), Peak Signal to Noise Ratio (PSNR), Structural Similarity Index Metric (SSIM), Mean Square Error (MSE), Root Mean Square Error (RMSE) and Image Enhancing Factor (IEF). Each of the eight filters will be applied to four types of medical images namely, X-rays, Computed Tomography (CT), Magnetic Resonance Imaging (MRI) and Ultrasound. These four types of medical images will be corrupted separately with eight different types of noises namely, Poisson, Gaussian, Speckle, Rayleigh, Uniform, Impulse, Local Variance and Salt and Pepper noise. The noise intensity of each type of noise will be varied from 25% to 70% using an interval of 15% when corrupting he medical images. The tests will be carried out in two types of image formats which are greyscale format and RGB format. In order to ensure that the results prove the filter’s capabilities, the experiment was carried out on a sample size of 200 medical images – 50 samples for each type of medical image. By analyzing the average performance of each filter on each medical image corrupted by a certain type of noise at varying noise intensity and testing them on two different image formats, a conclusion was derived on the type of filter that was most effective in denoising an image and most time efficient under each circumstance.