Scalable accelerator architecture for computing video quality metrics
A scalable hardware accelerator configured to compute video quality metrics is disclosed. In some embodiments, an accelerator for video quality metrics comprises an application-specific integrated circuit that includes a buffer memory configured to store at least a portion of a reference frame of a video and at least a corresponding portion of a distorted frame of a transcoded version of the video and that includes a processing unit configured to receive data from the buffer memory and compute a perception-based video quality metric for the distorted frame with respect to the reference frame.
1. A system, comprising:
a buffer memory of an application-specific integrated circuit configured to store at least a portion of a reference frame of a video and at least a corresponding portion of a distorted frame of a transcoded version of the video;
a programming interface of the application-specific integrated circuit configured to receive a selection of one of a plurality of supported full reference, perception-based objective video quality metrics; and
a processing unit of the application-specific integrated circuit configured to receive data from the buffer memory and simultaneously compute a plurality of video quality metrics in parallel including the selected one of the plurality of supported full reference, perception-based objective video quality metrics for the distorted frame with respect to the reference frame and a no reference video quality metric for the reference frame that indicates source quality prior to transcoding by determining pixel spread from computing pixel edge width values.
2. The system of claim 1 , further comprising a controller of the application-specific integrated circuit configured to obtain reference frame data and distorted frame data from a main memory that is external to the system.
3. The system of claim 1 , wherein the processing unit is further configured to scale received data from the buffer memory to a prescribed viewport resolution.
4. The system of claim 1 , wherein the processing unit is further configured to simultaneously compute a Peak Signal-to-Noise Ratio (PSNR) metric for the distorted frame with respect to the reference frame in parallel.
5. The system of claim 1 , wherein the no reference video quality metric comprises a blurriness metric of the reference frame.
6. The system of claim 1 , wherein the processing unit is configured to compute one or more perception-based video quality metrics.
7. The system of claim 6 , wherein the one or more perception-based video quality metrics comprises a Structural Similarity Index Measure (SSIM), a Multi-Scale SSIM (MS-SSIM), a Visual Information Fidelity (VIF), a Video Multimethod Assessment Fusion (VMAF), and a Detail Loss Metric (DLM).
8. The system of claim 1 , wherein the processing unit is configured to simultaneously compute three video quality metrics in parallel.
9. The system of claim 1 , wherein the processing unit is further configured to scale the received reference frame data using a first scaling unit and the received distorted frame data using a second scaling unit prior to computing the selected one of the plurality of supported full reference, perception-based objective video quality metrics for the distorted frame with respect to the reference frame.
10. The system of claim 1 , wherein the processing unit is configured to compute a plurality of video quality metrics in parallel in real time.
11. The system of claim 1 , wherein the processing unit is configured to compute a plurality of video quality metrics for different viewport resolutions.
12. The system of claim 1 , wherein the computed perception-based video quality metric comprises a frame level score, one or more block level scores, or both.
13. The system of claim 1 , wherein the system comprises the application-specific integrated circuit.
14. The system of claim 1 , wherein the system comprises a component of a video transcoding system.
15. A method, comprising:
receiving data from a buffer memory of an application-specific integrated circuit comprising at least a portion of a reference frame of a video and at least a corresponding portion of a distorted frame of a transcoded version of the video;
receiving from a programming interface of the application-specific integrated circuit a selection of one of a plurality of supported full reference, perception-based objective video quality metrics; and
simultaneously computing at a processing unit of the application-specific integrated circuit a plurality of video quality metrics in parallel including the selected one of the plurality of supported full reference, perception-based objective video quality metrics for the distorted frame with respect to the reference frame and a no reference video quality metric for the reference frame that indicates source quality prior to transcoding by determining pixel spread from computing pixel edge width values.
16. The method of claim 15 , wherein the processing unit is configured to simultaneously compute a Peak Signal-to-Noise Ratio (PSNR) metric for the distorted frame with respect to the reference frame in parallel.
17. The method of claim 15 , wherein the no reference video quality metric comprises a blurriness metric of the reference frame.
18. A computer program product embodied in a non-transitory computer readable medium and comprising computer instructions for:
receiving data from a buffer memory of an application-specific integrated circuit comprising at least a portion of a reference frame of a video and at least a corresponding portion of a distorted frame of a transcoded version of the video;
receiving from a programming interface of the application-specific integrated circuit a selection of one of a plurality of supported full reference, perception-based objective video quality metrics; and
simultaneously computing at a processing unit of the application-specific integrated circuit a plurality of video quality metrics in parallel including the selected one of the plurality of supported full reference, perception-based objective video quality metrics for the distorted frame with respect to the reference frame and a no reference video quality metric for the reference frame that indicates source quality prior to transcoding by determining pixel spread from computing pixel edge width values.
19. The computer program product of claim 18 , wherein the processing unit is configured to simultaneously compute a Peak Signal-to-Noise Ratio (PSNR) metric for the distorted frame with respect to the reference frame in parallel.
20. The computer program product of claim 18 , wherein the no reference video quality metric comprises a blurriness metric of the reference frame.