Level difference set based coarse to fine image representation and rendering

A nonlinear scale space representation of image is proposed and applied to time critical rendering. First of all, an image is decomposed into level difference sets and the connected components of each level difference set are measured. Then the connected components are ordered by size from large to small. The connected components are displayed in order of size when rendering. The level detail of the displayed image increases as time passes. The method is characterized by shaped invariance and contrast invariance. The method is also tested experimentally. Finally, the limit of the method is discussed and the improvement is given.