부산대학교 도서관

The quantum noise, intrinsic to the acquisition of X-ray images, becomes more significant during the processes related to fluoroscopy due to its low signal to noise ratio. In this condition, and if there is presence of motion in the image sequence, the use of temporal filtering is restricted, because it will lead to the generation of traces in the resulting filtered images. This artifact can be minimized using purely spatial filtering on regions where motion has been detected. The combination of temporal filtering / spatial filtering results in the hybrid filtering, which is driven by a binary mask of motion detection and it is obtained from the analysis and classification of pixels, based on statistical properties of the image. In this sense, this article aims at contributing to the methods of motion detection, presenting new analyzes in the sensibility of a detection algorithm and tests with real images of X-ray examinations. Preliminary results indicate that the filtering process can be improved with the proper choice of parameters of the motion detection mask for different types of medical X-ray procedures, which differ fundamentally by the amount of relative movement among consecutive frames and the contrast among gray intensities in an image sequence, in relation to the static background.