Temporally-Offset Sampling for High-Speed HDR Video
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
High-quality video capture is hindered by the need for expensive high-speed cameras and the challenges of motion aliasing and poor signal-to-noise ratio in high dynamic range (HDR) video, especially in portable devices where traditional cameras use constant exposure times and cannot capture enough frames per second.
Innovation Solution
The approach involves capturing temporally-consecutive image samples with varying exposure times for each pixel, offset relative to other pixels, generating synthetic samples by combining intensity, and grouping these samples to construct video frames, allowing for high-speed HDR video capture using a single low-speed camera by exploiting temporal and spatial redundancy.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Productivity
If high-speed cameras are used to capture high-quality video with high frames per second, then video quality and frame rate are improved, but device cost and complexity increase significantly
Solution Approach 1:
The camera sensor is divided into multiple pixel groups, where each group captures images at different exposure times. This segmentation allows the system to simulate high-speed capture by combining data from multiple slower captures, achieving high frame rate效果 without requiring a physically high-speed camera.
Solution Approach 2:
Multiple images are captured in advance at different exposure times before the final video frame is constructed. By pre-capturing multiple frames with varying exposures and then synthesizing them, the system achieves high-quality high-speed video without needing a high-speed camera hardware.
2Device complexity
If constant full-frame exposure time is used in traditional video cameras, then device cost and complexity are reduced, but frame rate and video quality deteriorate
Solution Approach 1:
The exposure time for each pixel group is made dynamic rather than constant. Different pixel groups use different exposure times that vary over time, allowing the system to capture multiple frames at different exposures and synthesize high frame rate video from a low-speed camera hardware.
3Manufacturing precision
If short exposure time is used to capture fast-moving objects, then motion aliasing is reduced, but light capture decreases resulting in poor signal-to-noise ratio
Solution Approach 1:
Multiple images captured at different exposure times are merged and combined to form the final video frame. By combining these images, the system achieves both sharp motion capture (from shorter exposures) and adequate light capture (from longer exposures), resolving the contradiction between motion accuracy and signal-to-noise ratio.
Data Source
AI summary
Various embodiments are directed to generating video data using temporally offset image data samples having disparate exposure times. Synthetic samples are generated for each of the pixels at a particular time period by computing, for each synthetic sample, a combined intensity of the captured samples that fall within the time period. Synthetic samples from adjacent pixels are grouped and image data in different groups is compared to identify matching groups. Video frames are constructed by combining image data from the captured samples based upon the matched images.