Border handling for motion compensated temporal interpolator using camera model

摘要

In one embodiment of the present invention, a method is provided for providing border handling in motion compensated interpolation, wherein a camera model is used to detect areas of predominant movements in a displayable output, and for each reveal or conceal area: it is determined whether the reveal or conceal areas is due to a camera movement or an object movement. If the reveal or conceal areas is due to a camera movement. The camera model is used to arrive at interpolated pixel values for pixels within the corresponding area. If the reveal or conceal areas is not due to a camera movement, arriving at interpolated pixel values for pixels within the corresponding area without using the camera model.

主权项

1. A method for providing border handling in motion compensated interpolation, the method comprising:
distinguishing between border areas and non-border areas of displayable output, the displayable output representing one frame of a sequence of frames, the border areas located at edges of the one frame of the sequence of frames; using a camera model to detect areas of predominant movements in the displayable output, wherein the camera model identifies probable pixels in motion in the border areas and re-generates a vector field for the pixels in probable motion in the border areas of the displayable output based on common camera movements; locating reveal areas and conceal areas near the border areas of the displayable output, wherein a reveal area is an area visible in a current frame that was not visible in a previous frame and a conceal area is an area visible in the previous frame that is not visible in the current frame; and for each reveal area or conceal area near a border of the displayable output:
applying a camera correlation heuristic to determine whether the reveal area or conceal area near the border of the displayable output is due to a camera movement or due to an object movement, the heuristic applied by comparing an actual vector field for the respective reveal area or conceal area to the vector field for the respective reveal area or conceal area re-generated by the camera model;based on a determination that the reveal area or conceal area is due to a camera movement, using the camera model to arrive at interpolated pixel values for pixels within the corresponding reveal or conceal area; andbased on a determination that the reveal area or conceal area is not due to a camera movement, arriving at interpolated pixel values for pixels within the corresponding reveal or conceal area without using the camera model.