METHOD AND SYSTEM FOR NOISE REDUCTION WITH A MOTION COMPENSATED TEMPORAL FILTER
Described herein is a method and system for the reduction of noise in a video sequence. When motion is present in the video sequence, this system and method identifies motion data. With the motion data, a Motion Compensated Temporal Filter (MCTF) can apply motion compensation prior to filtering in the time domain. Temporal filtering can be performed to reduce visible noise and other detrimental artifacts.
1 . A method for the reduction of noise in a current video block, said method comprising:
in a motion compensated temporal filter (MCTF),
generating for the current video block candidate motion vectors and respective confidence values,
wherein the confidence values provide an indication of accuracy of each of the candidate motion vectors in a prediction process involving a reference block;
selecting one of the candidate motion vectors based on an evaluation of the confidence values, the selected motion vector associated with a first confidence value;
generating a first motion compensated block though application of the selected motion vector to a first reference block; and
filtering the current video block and the first motion compensated block according to the first confidence value to generate a current filtered block, the filtering comprising:
scaling the current video block according to a first scaling value;
scaling the first motion compensated block according to a second scaling value; and
combining the scaled current video block with the scaled first motion compensated block, wherein the first and second scaling values are based on the first confidence value.