Back to results
US 9,883,083 B2Granted
Processing prior temporally-matched frames in 3D-based video denoising
AssigneeCisco Technology, Inc.
InventorsJoel W. Schoenblum, Larry S. McKinney, Kenneth W. Heinrich
Life of this patent
Estimating the term…
Abstract
In one system embodiment, an overlapped block processing module configured to provide three-dimensional (3D) denoising of plural frames corresponding to a raw video sequence; and a frame alignment module configured to represent the raw video sequence with motion compensated frames corresponding to the raw video sequence, the motion compensated frames consisting of the plural frames and fewer in quantity than the quantity of frames of the raw video sequence, the plural frames based on prior temporally matched frames corresponding to the raw video sequence.
Claims20
A frame alignment module, comprising:
- a memory;
- one or more processors configured to execute instructions stored in the memory, the instructions comprising:
- summing a multiplied frame with a first delayed frame and a reference frame to provide a summation frame, wherein the first delayed frame comprises an estimate of at least one of a plurality of first plural frames from another of the plurality of first plural frames;
- delaying the summation frame to provide a delayed summation frame wherein delaying comprises an imposition of a defined delay that varies based on an inputted frame;
- and
- frame matching the delayed summation frame with the reference frame to provide a second summation frame, wherein the frame matching comprises processing of luminance and chrominance data of the plurality of first plural frames.
3 more elements
A frame alignment module, comprising:
- a memory;
- one or more processors configured to execute instructions stored in the memory, the instructions comprising:
- providing a summation frame comprised by summing a multiplied frame with a first delayed frame and a reference frame to provide the summation frame;
- delaying the summation frame by the frame alignment module to provide a delayed summation frame wherein delaying comprises an imposition of a defined delay that varies based on an inputted frame;
- and
- frame matching the delayed summation frame by the frame alignment module with the reference frame to provide a second summation frame, wherein the frame matching comprises processing of luminance and chrominance data of inputted frames, wherein one or more outputted frames from the second summation frame comprise an approximation of a received raw video sequence, wherein frames of the received raw video sequence number greater in quantity than the one or more outputted frames.
3 more elements
A system, comprising:
- a memory;
- one or more processors configured to execute instructions stored in the memory, the instructions comprising:
- receiving a first temporal sequence of first plural frames and a reference frame, at least a portion of the first plural frames and the reference frame corrupted by noise;
- frame matching the first plural frames;
- summing the frame-matched first plural frames with at least one of the first plural frames to provide a first summation frame;
- delaying the frame-matched first plural frames to provide a first delayed frame;
- frame matching the first summation frame to the reference frame to provide a second summation frame;
- self-combining the second summation frame multiple times to provide a multiplied frame;
- summing the multiplied frame with the first delayed frame and the reference frame to provide a third summation frame;
- delaying the third summation frame to provide a delayed third summation frame, wherein each of the delaying comprises an imposition of a defined delay that varies based on the inputted frame;
- frame matching the delayed third summation frame with the reference frame to provide a fourth summation frame, wherein each of the frame matching comprises processing of luminance and chrominance data of the inputted frames;
- and
- outputting the first delayed frame, the second summation frame, the reference frame, and the fourth summation frame for further processing.
10 more elements
Dependencies are inferred from the claim text.