Hybrid memory compression scheme for decoder bandwidth reduction
View Patent ↗A method for reducing memory bandwidth in a video decoder begins by performing a data reduction operation on a decoded first coded image to produce a second set of image data. The second set of image data stored and is selectively used for subsequent image decoding, thereby reducing the memory bandwidth. The data reduction operation can include image downsampling, wherein the pixel density is reduced by a factor of two in each of the vertical and horizontal directions.
1. A method for performing motion compensated prediction for video images, comprising the steps of:
performing a data reduction operation on a decoded image to produce a set of reduced image data;
storing non-reduced image data and the set of reduced image data;
receiving a motion vector;
comparing a measure of a size of the motion vector to a predetermined threshold;
conditionally selecting between the stored non-reduced image data and the stored set of reduced image data based on the result of the comparison; and
performing motion compensated prediction using the motion vector and the selected image data.
2. The method according to claim 1 , wherein the non-reduced image data is a full resolution image.
3. The method according to claim 1 , wherein the data reduction operation is image downsampling.
4. The method according to claim 3 , wherein the image downsampling reduces pixel density by a factor of two in each of the vertical and horizontal directions.
5. The method according to claim 1 , wherein the predetermined threshold is eight pixels.
6. The method according to claim 1 , wherein the selecting step includes selecting the reduced set of image data if the measure of the size of the motion vector is greater than the predetermined threshold.
7. The method according to claim 1 , wherein the selecting step includes selecting the non-reduced image data if the measure of the size of the motion vector is less than the predetermined threshold.
8. A video decoder configured to perform motion compensated prediction, comprising:
an image downsampler operable to downsample a decoded image to form reduced image data;
a memory operable to store non-reduced image data and the reduced image data;
a motion vector comparator operable to compare a measure of a size of a motion vector to a predetermined threshold;
an image selector operable to conditionally select between the stored reduced image data and non-reduced image data based on the result of the comparison to produce selected image data; and
a motion predictor operable to perform motion compensated prediction using the motion vector and the selected image data.
9. The video decoder according to claim 8 , wherein the non-reduced image data is a full resolution image.
10. The video decoder according to claim 8 , wherein the image downsampler is operable to reduce pixel density by a factor of two in each of the vertical and horizontal directions.
11. The video decoder according to claim 8 , wherein the predetermined threshold is eight pixels.
12. The video decoder according to claim 8 , wherein the selector is operable to select the reduced set of image data if the measure of the size of the motion vector is greater than the predetermined threshold.
13. The video decoder according to claim 8 , wherein the selector is operable to select the non-reduced image data if the measure of the size of the motion vector is less than the predetermined threshold.