Methods and systems for mixed spatial resolution video compression
A method for encoding a frame of visual data which includes the steps of encoding an original full resolution frame, storing coded data for the encoded full resolution frame, reconstructing and storing encoded full resolution frame, downsampling the original full resolution frame to render it a reduced spatial resolution frame, encoding the reduced spatial resolution frame, storing coded data for the reduced spatial resolution frame, reconstructing and storing the reduced spatial resolution frame, upsampling and storing the reconstructed reduced spatial resolution frame, comparing a characteristic in the reconstructed full resolution frame with said characteristic in the original full resolution frame to determine the deviation of the reconstructed full resolution frame from the original full resolution frame with respect to said characteristic, comparing said characteristic in the upsampled reconstructed spatial reduced resolution frame with said characteristic in the original full resolution frame to determine the deviation of the upsampled reconstructed spatial reduced resolution frame from the original full resolution frame with respect to said characteristic, selecting the frame with the lesser deviation from the original full resolution frame with respect to said characteristic, and outputting the coded data corresponding to the frame with the lesser deviation from the original full resolution frame with respect to said characteristic to the bitstream.
1. A computer implemented method of video compression, the method comprising the steps of:
generating a reconstructed full resolution frame and coded data corresponding to the reconstructed full resolution frame from an original full resolution frame;
generating an upsampled reconstructed reduced spatial resolution frame and coded data corresponding to the upsampled reconstructed reduced spatial resolution frame from the original full resolution frame;
determining a first deviation by comparing at least one characteristic in the reconstructed full resolution frame with said at least one characteristic in the original full resolution frame;
determining a second deviation by comparing said at least one characteristic in the upsampled reconstructed reduced spatial reduced resolution frame with said at least one characteristic in the original full resolution frame; and
outputting the coded data corresponding to the reconstructed full resolution frame when the first deviation is less than the second deviation; otherwise
outputting the coded data corresponding to the upsampled reconstructed reduced spatial reduced resolution frame.
2. The method of claim 1 , wherein generating the reconstructed full resolution frame and coded data corresponding to the reconstructed full resolution frame includes the steps of:
encoding the original full resolution frame to generate an encoded full resolution frame,
storing coded data for the encoded full resolution frame, and
reconstructing and storing the encoded full resolution frame.
3. The method of claim 1 , wherein generating the reconstructed reduced spatial resolution frame and coded data corresponding to the reconstructed reduced spatial resolution frame comprises the steps of:
downsampling the original full resolution frame to render the original full resolution frame into a reduced spatial resolution frame,
encoding the reduced spatial resolution frame,
storing coded data for the reduced spatial resolution frame,
reconstructing and storing the reduced spatial resolution frame, and
upsampling and storing the reconstructed reduced spatial resolution frame.
4. The method of claim 2 , wherein the step of encoding the original full resolution frame further comprises the step of applying discrete cosine transform-based downsampling to the original full resolution frame.
5. The method of claim 3 , wherein the step of encoding the reduced spatial resolution frame further comprises the step of applying discrete cosine transform-based upsampling to the reduced spatial resolution frame for encoding.
6. The method of claim 3 , wherein the step of upsampling and storing the reconstructed reduced spatial resolution frame further comprises the step of applying tweening-based spatial interpolation to the reconstructed reduced spatial resolution frame.
7. The method of claim 1 , wherein the coded data corresponding to the reconstructed full resolution frame and the coded data corresponding to the upsampled reconstructed reduced spatial resolution frame are each stored in a memory buffer.
8. The method of claim 1 , wherein the first deviation and the second deviation are each determined based on a visual-perceptual metric.
9. The method of claim 8 , wherein the visual-perceptual metric is a just noticeable difference metric.
10. A system for video compression, comprising:
a full resolution encoder for generating a reconstructed full resolution frame and coded data corresponding to the reconstructed full resolution frame from an original full resolution frame;
a reduced resolution encoder for generating an upsampled reconstructed reduced spatial resolution frame and coded data corresponding to the upsampled reconstructed reduced spatial resolution frame from the original full resolution frame; and
a decision component for:
determining a first deviation by comparing at least one characteristic in the reconstructed full resolution frame with said at least one characteristic in the original full resolution frame
determining a second deviation by comparing said at least one characteristic in the upsampled reconstructed reduced spatial reduced resolution frame with said at least one characteristic in the original full resolution frame; and
outputting the coded data corresponding to the reconstructed full resolution frame when the first deviation is less than the second deviation; otherwise
outputting the coded data corresponding to the upsampled reconstructed reduce spatial reduced resolution frame.
11. The system of claim 10 , wherein generating the reconstructed full resolution frame and coded data corresponding to the reconstructed full resolution frame includes the steps of:
encoding the original full resolution frame to generate an encoded full resolution frame,
storing coded data for the encoded full resolution frame, and
reconstructing and storing the encoded full resolution frame.
12. The system of claim 10 , wherein generating the reconstructed reduced spatial resolution frame and coded data corresponding to the reconstructed reduced spatial resolution frame comprises the steps of:
downsampling the original full resolution frame to render the original full resolution frame into a reduced spatial resolution frame,
encoding the reduced spatial resolution frame, storing coded data for the reduced spatial resolution frame,
reconstructing and storing the reduced spatial resolution frame, and
upsampling and storing the reconstructed reduced spatial resolution frame.
13. The system of claim 11 , wherein encoding the original full resolution frame further comprises applying discrete cosine transform-based downsampling to the original full resolution frame is.
14. The system of claim 12 , wherein encoding the reduced spatial resolution frame further comprises applying discrete cosine transform-based upsampling to the reduced spatial resolution frame.
15. The system of claim 12 , wherein upsampling and storing the reconstructed reduced spatial resolution frame further comprises applying tweening-based spatial interpolation to the reconstructed reduced spatial resolution frame.
16. The system of claim 10 , wherein the coded data corresponding to the reconstructed full resolution frame and the coded data corresponding to the upsampled reconstructed reduced spatial resolution frame are each stored in a memory buffer.
17. The system of claim 10 , wherein the first deviation and the second deviation are each determined based on a visual-perceptual metric.
18. The system of claim 17 , wherein the visual-perceptual metric is a just noticeable difference metric.