Blooming artifact reduction for cardiac CT

With modern cardiac CT scanners detection and classification of coronary artery diseases is a common task. More and more low dose cardiac CT prevails the clinical routine. To suppress noise, a smooth reconstruction kernel is used. Unfortunately this inducts an additional blurring to the images. At high contrast graduations this so-called blooming makes it hard to identify small structures especially inside vessels. The lumen inside the vessel seems to be much smaller than it actually is. We propose a new image based algorithm which can be easily integrated in existing systems. We use two standard reconstructed images one with low and one with high resolution as basis. Our algorithm identifies the image regions with high contrast graduations by performing a median filter with an additional calculation of the local variance. The high contrast regions will be replaced by a bilateral filtered high resolution image whereas the low contrast region will be left at low resolution by using the low resolution image. The final image, including its noise texture, looks like a standard cardiac CT image with the exception of reduced blooming artifacts.