Back to results
US 9,317,958 B2Granted
Auto-convergence system with active learning and related method and machine-readable medium thereof
AssigneeMediatek Inc.
InventorChia-Ming Cheng
Life of this patent
Estimating the term…
Abstract
An auto-convergence system includes a disparity unit, a convergence unit and an active learning unit. The disparity unit performs a disparity analysis upon an input stereo image pair, and accordingly obtains a disparity distribution of the input stereo image pair. The convergence unit adjusts the input stereo image pair adaptively according to the disparity distribution and a learned convergence range, and accordingly generates an output stereo image pair for playback. The active learning unit actively learns a convergence range during playback of stereo image pairs, and accordingly determines the learned convergence range.
Claims18
An auto-convergence system, comprising:
- a disparity unit, arranged for performing a disparity analysis upon an input stereo image pair, and accordingly obtaining a disparity distribution of the input stereo image pair, wherein the disparity distribution is obtained based on disparity values of the input stereo image pair, and the disparity distribution is within an original disparity range delimited by a largest disparity value and a smallest disparity value among the disparity values of the input stereo image pair;
- a convergence unit, coupled to the disparity unit and arranged for deriving a specific disparity range from the original disparity range, adaptively adjusting the input stereo image pair according to a comparison result derived from comparing the specific disparity range with a learned convergence range, and accordingly generating an output stereo image pair for playback;
- and
- an active learning unit, coupled to the convergence unit and arranged for actively learning a convergence range during playback of stereo image pairs and accordingly determining the learned convergence range;
- wherein the specific disparity range is delimited by a first disparity value and a second disparity value, and the first disparity value is larger than the second disparity value, and
- wherein the convergence range is a disparity range delimited by a third disparity value and a fourth disparity value, and the third disparity value is larger than the fourth disparity value.
3 more elements
An auto-convergence method, comprising:
- performing a disparity analysis upon an input stereo image pair, and accordingly obtaining a disparity distribution of the input stereo image pair, wherein the disparity distribution is obtained based on disparity values of the input stereo image pair, and the disparity distribution is within an original disparity range delimited by a largest disparity value and a smallest disparity value among the disparity values of the input stereo image pair;
- actively learning a convergence range during playback of stereo image pairs, and accordingly determining a learned convergence range;
- and
- performing an active learning process for deriving a specific disparity range from the original disparity range, adaptively adjusting the input stereo image pair according to a comparison result derived from comparing the specific disparity range with the learned convergence range, and accordingly generating an output stereo image pair for playback;
- wherein the specific disparity range is delimited by a first disparity value and a second disparity value, and the first disparity value is larger than the second disparity value;
- and
- wherein the convergence range is a disparity range delimited by a third disparity value and a fourth disparity value, and the third disparity value is larger than the fourth disparity value.
4 more elements
A non-transitory machine-readable medium storing a program code, wherein when the program code is executed by a processor, the processor is instructed to perform following steps:
- performing a disparity analysis upon an input stereo image pair, and accordingly obtaining a disparity distribution of the input stereo image pair, wherein the disparity distribution is obtained based on disparity values of the input stereo image pair, and the disparity distribution is within an original disparity range delimited by a largest disparity value and a smallest disparity value among the disparity values of the input stereo image pair;
- actively learning a convergence range during playback of stereo image pairs and accordingly determining a learned convergence range;
- and
- performing an active learning process for deriving a specific disparity range from the original disparity range, adaptively adjusting the input stereo image pair according to a comparison result derived from comparing the specific disparity range with and the learned convergence range, and accordingly generating an output stereo image pair for playback;
- wherein the specific disparity range is delimited by a first disparity value and a second disparity value, and the first disparity value is larger than the second disparity value;
- and
- wherein the convergence range is a disparity range delimited by a third disparity value and a fourth disparity value, and the third disparity value is larger than the fourth disparity value.
4 more elements
Dependencies are inferred from the claim text.