Methods and systems for determining data-adaptive weights for motion estimation in a video sequence
Aspects of the present invention are related to systems and methods for determining local-analysis-window size and weighting parameters in a gradient-based motion estimation system.
1. A method for determining data-adaptive weights in a motion
estimation system, said method comprising:
selecting, in a computing system associated with a motion estimation system, a size of a local analysis window based on a local structural variation in a current image and a noise estimate associated with said current image;
receiving a plurality of image pixel values associated with said local analysis window in said current image;
identifying a window-center image pixel value associated with a window center location in said local analysis window;
calculating a weight associated with a first location in said local analysis window based on a difference between said window-center image pixel value and an image pixel value associated with said first location; and
performing gradient-based motion estimation using said weight to estimate a motion vector at said window center location.
2. The method as described in claim 1 , wherein said weight associated with said first location is a first value when said difference and a threshold meet a first criterion and a second value when said difference and said threshold do not meet said first criterion.
3. The method as described in claim 1 , wherein said weight associated with said first location is based on an exponential function dependent on said difference.