Implementation of a selective mean filter algorithm to reduce noise in computed radiography images

Abstract

This study implements a selective mean filter (SMF) and determines its effectiveness on computed radiography (CR) images. A phantom of the Kualitas Citra Imajing (KUCING) was used. The effectiveness of the SMF was compared to the adaptive mean filter (AMF). The assessment of image quality was based on several parameters, such as the peak signal to noise ratio (PSNR), mean squared error (MSE), structural similarity index (SSIM), contrast consistency (CV), contrast linearity (CL), and modulation transfer function (MTF). The results show that the PSNR value generated by the SMF is higher than with the AMF. The MSE value generated by the SMF algorithm is lower than AMF. The SSIM values showed that SMF algorithm produces a better image similarity between the original image and the filtered image compared to the AMF algorithm. In consistency and linearity, the AMF leads to significant changes from the original image, while the SMF image is closer to the original image. The 10% MTF generated by the SMF is the same as the original image, while AMF produces a significant decrease in the 10% MTF. The SMF took 17.9 ± 0.7 seconds for denoising one CR image, while AMF took 519 ± 5 seconds. SMF decreases image noise without a significant effect on the spatial resolution of the image. SMF filtering is superior to AMF filtering in terms of spatial resolution and a shorter computational time.