Efficient and Robust Filtering Method for Medical CT Images

This paper introduces a new approach to ensure the certainty of medical diagnosis by eliminating the salt-and-pepper noise (SPN) in medical applications for both gray and coloured computed tomography (CT) images. The proposed approach is based on median filter which utilized for value-preserving and edge-preserving in digital image processing applications, Thus, the proposed approach is called improved adaptive median filter (IAMF). In contrast to the available research in the literature, the introduced method is characterized by the high filtering quality, robust in different noise intensities (low, medium and high) and computed efficiently. The obtained results of the filtering process have been analysed in terms of four main metrics: peak signal to noise ratio (PSNR), structure similarity (SSIM), universal image quality (UIQ) and filter average execution time (AET). The test scenarios were conducted using MATLAB 2019a running on Windows 10 computer. The success of the proposed filter has been validated by the statistical analysis based on the aforementioned metrics using gray and coloured medical CT images. For the worst case scenario in gray and coloured CT images when the noise intensity is 90%, the IAMF enhances the PSNR by 108-129%, the SSIM by 105-153% and the UIQ by 97-100% when they were compared to different filters that existed in the recent literature. These ratios depend on the image quality and image resolution. Moreover, the filter execution time has been improved by five times in gray scenarios and four times in coloured scenarios. Finally, the obtained results are visually verified as well.