Noise reduction and segmentation in time-varying ultrasound images

Steps in a procedure for the automatic measurement of the cardiac ejection fraction in time-varying two-dimensional ultrasound images are presented. Two statistical contextual noise reduction methods are generalized to handle noisy three-dimensional gray-level images (the third dimension being time). The improved gray-level images are thresholded by an adaptive thresholding algorithm. The concept of connected components in, three-dimensional binary images is introduced and used to segment structures from noise according to their extension in time. It is concluded that the automatic determination of the cardiac ejection fraction in ultrasound images will be possible