Method for pixel prediction with low complexity
View Patent ↗A method including: calculating a prediction of a first block of pixels in a second video frame with a temporal position between a first and a third video frame from a first motion vector of pixel position components relative to spatial pixel positions of the first block referring the first video frame and a third motion vector of pixel position components relative to spatial pixel positions of the first block referring the third video frame, wherein a second block of pixels in the third video frame spatially collocated with the first block is predicted from a second motion vector of pixel position components relative to spatial pixel positions of the second block referring the first video frame.
1. A method comprising:
calculating a prediction of a first block of pixels in a second video frame with a temporal position between a first and a third video frame from a first motion vector of pixel position components relative to spatial pixel positions of the first block referring to the first video frame and a third motion vector of pixel position components relative to spatial pixel positions of the first block referring the third video frame, wherein a second block of pixels in the third video frame spatially collocated with the first block is predicted from a second motion vector of pixel position components relative to spatial pixel positions of the second block referring the first video frame,
wherein said calculating the prediction of the first block of pixels in the second video frame includes
calculating the first and the third motion vectors from respective downscalings of said second motion vector according to the temporal position of the second video frame between the first and the third video frame so that the first motion vector equals the third motion vector plus the second motion vector,
deriving a corrected third motion vector by rounding the pixel position components of the third motion vector to nearest integer values,
defining a first error vector as a difference between the third motion vector and the corrected third motion vector,
deriving a corrected first motion vector by subtracting the error vector from the first motion vector and quantifying according to a defined pixel position resolution of the first video frame, and
averaging pixels of a block-sized area of the first frame to which the corrected first motion vector refers with corresponding pixels of a block-sized area of the third frame to which the corrected third motion vector refers.
2. The method according to claim 1 , wherein
the temporal position of the second video frame is halfway between the first and the third video frames, and the step of deriving the corrected first motion vector includes defining the corrected first motion vector to be opposite of the corrected third motion vector.