[Analysis of the frequency components of CR portal images and application to multi-objective frequency processing]

The image contrast of portal images obtained using computed radiography (CR) is low for radiation therapy, and their noise level is high. In order to improve the image quality of CR portal images, we investigated the relationship between the frequency components of the bone edge and noise area. We used a personal computer, conventional flatbed scanner, and public-domain software (NIH Image). To analyze the frequency components of bone edges, we generated difference images from the original and unsharp mask images of various sizes. In addition, we analyzed the frequency components of noise, after changing the cut-off frequency on the power spectrum to process a low- and high-pass filter. We found that the frequency components of bone edges and noise areas ranged from 0.18 to 0.25 cycle/mm and 1.30 cycles/mm, respectively. We conclude that this process can remove noise from homogeneous areas and preserve bone edges using multi-objective frequency processing.