Comparison of Noise Removal Methods for Noisy Lenke Type 1 Idiopathic Scoliosis Images

Improving the quality of medical images is essential for accurate clinical diagnosis. However, medical images are often plagued by various types of noise, posing challenges to the reliability and accuracy of image diagnosis. This study aims to compare Anisotropic, Wiener, Median, Average, and Gaussian filters on noisy images to improve the quality and accuracy of diagnostic results. To better adapt to the complex features in Lenke images, the contrast optimization algorithm was enhanced using Contrast-Limited Adaptive Histogram Equalization (CLAHE). In this study, artificial noise modeling (Gaussian and Quantum noise) was applied to Lenke X-ray images, and a comparative analysis was performed to discover the most effective filtering scheme for a particular type of noise based on various parameters such as MSE, PSNR, SNR, and SSIM. The results unveiled that the Wiener filter was relatively effective for both types of noise. The Gaussian filter also performed satisfactorily. Nonetheless, no filter could be declared the most effective for both types of noise. However, their performance varied according to the type of noise present and the evaluation parameters.