관상동맥 조영영상에서 라플라시안 필터와 고속 비지역적 평균 알고리즘의 유용성 평가

Abstract

This study aimed to reduce noise and blur in X-ray images acquired during coronary angiography by combining the Laplacian filter with the fast non-local means (FNLM) algorithm. The proposed method applies the Laplacian filter to im- ages and combines the filtered images with the original ones. Subsequently, the smoothing factor of the FNLM algorithm was adjusted, increasing from 0.001 to 0.1 in increments of 0.001, resulting in the generation of 100 denoised images for each image. The FNLM algorithm was optimized using the coefficient of variation (COV) and the contrast-to-noise ratio (CNR) as quantitative evaluation metrics. As a result, the COV gradually decreased with an increase in the smoothing factor, while the CNR increased; however, in certain images, the CNR decreased after reaching a specific threshold. Based on these findings, the smoothing factor was optimized to 0.051, which was identified as the inflection point for the CNR. The images processed with the optimized FNLM algorithm showed improvements of approximately 3.0 and about 3.5 times in COV and CNR, in comparison to the original images. Furthermore, an analysis of the differences in pixel values confirmed that edge enhancement in images processed with the optimized FNLM algorithm improved by an average of approximately 44.0% compared to the original images. In conclusion, when the Laplacian filter and the optimized FNLM algorithm significantly enhanced image quality while enhancing the edge.