Medical ultrasound image similarity measurement by human visual system (HVS) modeling
C Wallraven · Research Portal (King's College London) · 2002
We applied a four stage model of human vision to the problem of similarity measurement of medical (liver) ultrasound images. The results showed that when comparing a given image to a set that contained images with similar features, the model was able to correctly identify the most similar image in the set. Additionally, the shape of the similarity function closely followed a subjective measure of visual similarity for images around the best match. Removing some computational steps to reduce processing time enabled the comparison method to run in near real-time ([left angle bracket] 5 seconds), but with some acceptable loss of accuracy. These results could not be achieved using conventional similarity measurements based on image grey level statistics. (26 References).