Noise reduction for image sequences
View Patent ↗Described herein are techniques related to noise reduction for image sequences or videos. This Abstract is submitted with the understanding that it will not be used to interpret or limit the scope and meaning of the claims. A noise reduction tool includes a motion estimator configured to estimated motion in the video, a noise spectrum estimator configured to estimate noise in the video, a shot detector configured to trigger the noise estimation process, a noise spectrum validator configured to validate the estimated noise spectrum, and a noise reducer to reduce noise in the video using the estimated noise spectrum.
1. A tool to reduce noise in an image sequence, the tool comprising:
a motion estimator configured to yield a sequence of motion-compensated frames;
a noise spectrum estimator configured to estimate noise in one or more blocks of pixels in the sequence of input frames by:
identifying noisy and flat areas within the sequence of input frames; and
calculating an estimated noise power spectrum for a current frame in the sequence of input frames from the noisy and flat areas within the sequence of input frames;
a shot detector configured to detect a shot change within the sequence of input frames and to trigger the noise spectrum estimator in response to the detected shot change;
a noise reducer configured to reduce noise in the image sequence using the estimated noise power spectrum and the sequence of motion-compensated frames.
2. A tool according to claim 1 , wherein the noise spectrum estimator is configured to estimate noise using a pre-calculated estimated noise power spectrum.
3. A tool according to claim 1 , wherein the noise spectrum estimator is configured to estimate noise using a discrete cosine transform.
4. A tool according to claim 1 , wherein the noise spectrum estimator includes a Hamming window or a triangular window.
5. A tool according to claim 1 , wherein the noise spectrum estimator uses gradient analysis to locate blocks in the image sequence that contain no original texture.
6. A tool according to claim 1 , further comprising a noise spectrum validator configured to determine whether the estimated noise power spectrum is symmetrical.
7. A tool according to claim 1 , wherein the noise reducer includes a Wiener filter configured to reduce the noise in the image sequence.
8. A tool according to claim 1 , wherein the noise spectrum estimator is configured to estimate noise using a three-dimensional fast Fourier transform.
9. A tool according to claim 1 , wherein the noise spectrum estimator includes a Gaussian blurring filter configured to locate blocks in the image sequence that have flat areas and noise.
10. A tool according to claim 1 , wherein the noise reducer estimates a linear plane fit to a block of pixels in the sequence of input frames and to a block of pixels in the sequence of motion-compensated frames, removes a two-dimensional plane from the block of pixels in the sequence of input frames, and returns the removed two-dimensional plane to the block of pixels in the sequence of input frames.