Subjective distortion metric for rate distortion determinations
A video encoding system is configured to determine, based on image data, a distortion metric for a prediction mode candidate based on 1) gradient similarity between a prediction block of the image data and a reconstructed prediction block of reconstructed image data generated based on the image data and 2) motion between two frames of the image data. The video encoding system is also configured to select, from a plurality of prediction mode candidates, the prediction mode candidate as a prediction mode based at least in part on the distortion metric. Additionally, the video encoding system is configured to encode the image data based on the prediction mode.
1 . A video encoding system, configured to:
determine, based on image data, a distortion metric that is associated with a difference between the image data and reconstructed image data for a prediction mode candidate based on:
gradient similarity between a prediction block of the image data and a reconstructed prediction block of the reconstructed image data generated based on the image data; and
motion between two frames of the image data;
select, from a plurality of prediction mode candidates, the prediction mode candidate as a prediction mode based at least in part on the distortion metric; and
encode the image data based on the prediction mode.
2 . The video encoding system of claim 1 , wherein the video encoding system is configured to determine the gradient similarity based on:
a first absolute difference of a first magnitude of a horizontal filter applied to the prediction block and the reconstructed prediction block; and
a second absolute difference of a second magnitude of a vertical filter applied to the prediction block and the reconstructed prediction block.
3 . The video encoding system of claim 1 , wherein the video encoding system is configured to determine the motion between the two frames of the image data based on a motion vector indicative of a difference between a first frame of the two frames and a second frame of the two frames.
4 . The video encoding system of claim 3 , wherein the video encoding system is configured to:
determine a rate metric for the prediction mode candidate;
determine a rate-distortion metric based on the rate metric, the distortion metric, and the motion; and
select the prediction mode candidate based on the rate-distortion metric.
5 . The video encoding system of claim 4 , wherein the video encoding system is configured to determine the rate-distortion metric by weighting the rate metric relatively more heavily than the distortion metric the relatively higher the weight is.
6 . The video encoding system of claim 1 , wherein the image data comprises image data captured by an image sensor, downscaled image data, and full-resolution image data, wherein the downscaled image data and full-resolution image data are generated from the image data captured by the image sensor; and
the video encoding system is configured to encode the image data by encoding the full-resolution image data based on the prediction mode.
7 . The video encoding system of claim 1 , wherein the prediction mode candidate comprises a candidate inter prediction mode or a candidate intra prediction mode.
8 . The video encoding system of claim 1 , wherein the video encoding system is included in an electronic device.
9 . The video encoding system of claim 1 , wherein the video encoding system is configured to determine the distortion metric without determining a sum of squared difference.
10 . A non-transitory computer-readable medium comprising instructions that, when executed by processing circuitry, cause the processing circuitry to:
determine, based on image data, a distortion metric that is associated with a difference between the image data and reconstructed image data for a prediction mode candidate based on:
gradient similarity between a prediction block of the image data and a reconstructed prediction block of the reconstructed image data generated based on the image data; and
motion between two frames of the image data;
select, from a plurality of prediction mode candidates, the prediction mode candidate as a prediction mode based at least in part on the distortion metric; and
encode the image data based on the prediction mode.
11 . The non-transitory computer-readable medium of claim 10 , wherein the instructions, when executed, cause the processing circuitry to determine the gradient similarity based on:
a first absolute difference of a first magnitude of a horizontal filter applied to the prediction block and the reconstructed prediction block; and
a second absolute difference of a second magnitude of a vertical filter applied to the prediction block and the reconstructed prediction block.
12 . The non-transitory computer-readable medium of claim 11 , wherein the instructions, when executed, cause the processing circuitry to determine the motion between the two frames of the image data based on a motion vector indicative of a difference between a first frame of the two frames and a second frame of the two frames.
13 . The non-transitory computer-readable medium of claim 12 , wherein the instructions, when executed, cause the processing circuitry to determine the motion in:
a first manner for a first type of prediction mode candidate; and
a second manner different than the first manner for a second type of prediction mode candidate.
14 . The non-transitory computer-readable medium of claim 13 , wherein:
the first type of prediction mode candidate comprises an intra prediction mode candidate; and
the second type of prediction mode candidate comprises an inter prediction mode candidate.
15 . The non-transitory computer-readable medium of claim 10 , wherein the instructions, when executed, cause the processing circuitry to determine the distortion metric without determining a sum of squared difference.
16 . The non-transitory computer-readable medium of claim 15 , wherein the instructions, when executed, cause the processing circuitry to:
determine a rate-distortion metric based on the distortion metric; and
select the prediction mode candidate based on the rate-distortion metric.
17 . A computer-implemented method, comprising:
determining, via processing circuitry and based on image data, a distortion metric that is associated with a difference between the image data and the reconstructed image data for a prediction mode candidate based on:
gradient similarity between a prediction block of the image data and a reconstructed prediction block of the reconstructed image data generated based on the image data; and
motion between two frames of the image data;
selecting, via the processing circuitry and from a plurality of prediction mode candidates, the prediction mode candidate as a prediction mode based at least in part on the distortion metric; and
encoding, via the processing circuitry, the image data based on the prediction mode.
18 . The computer-implemented method of claim 17 , comprising:
determining, via the processing circuitry, the gradient similarity based on:
a first absolute difference of a first magnitude of a horizontal filter applied to the prediction block and the reconstructed prediction block; and
a second absolute difference of a second magnitude of a vertical filter applied to the prediction block and the reconstructed prediction block; and
determining, via the processing circuitry, the motion between the two frames of the image data based on a motion vector indicative of a difference between a first frame of the two frames and a second frame of the two frames.
19 . The computer-implemented method of claim 18 , wherein the horizontal filter and the vertical filter comprise Sobel filters.
20 . The computer-implemented method of claim 17 , comprising accounting, based on the gradient similarity, for one or more discontinuities introduced to the image data.