Camera Array Disparity Estimation for High Dynamic Range Imaging
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Conventional camera arrays struggle to capture high dynamic range images effectively due to regions of under- and over-saturation caused by varying exposure parameters across cameras, leading to poor disparity estimation and computational inefficiency in image processing.
Innovation Solution
A non-iterative method for estimating disparity using a camera array with varying exposure parameters, involving image rectification and a noise model to normalize images and compute weighted sums of shifted patches, which avoids saturation issues and reduces computational complexity.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Illumination intensity
If conventional camera arrays use varying exposure parameters across cameras, then dynamic range coverage is improved, but regions of under- and over-saturation occur leading to poor disparity estimation
Solution Approach 1:
The patent applies parameter changes by adjusting exposure parameters across different cameras in the array to capture a wider dynamic range. Each camera uses different exposure settings (e.g., different exposure times or ISO values) to capture both bright and dark regions of the scene, thereby improving overall dynamic range coverage while managing saturation through controlled parameter variation.
Solution Approach 2:
The patent introduces an intermediary processing step that estimates saturation levels and uses this information to guide disparity estimation. By incorporating saturation estimation as an intermediary mechanism, the system can identify reliable pixel regions for disparity calculation even when some cameras produce saturated images, thus maintaining measurement precision across varying exposure conditions.
2Illumination intensity
If conventional methods process images from camera arrays with varying exposures, then dynamic range is captured, but computational complexity and processing time increase
Solution Approach 1:
The patent applies preliminary action by pre-estimating saturation levels for each pixel across all cameras before performing disparity estimation. This preliminary saturation assessment allows the system to pre-identify reliable pixel regions and plan the disparity computation more efficiently, avoiding unnecessary calculations on saturated pixels and thereby reducing overall computational processing time.
Solution Approach 2:
The patent extracts and processes saturation information separately from the main disparity estimation pipeline. By taking out the saturation estimation as a distinct preprocessing step and using it to guide subsequent processing, the system avoids the computational burden of full-image processing and focuses computational resources only on reliable, non-saturated regions.
3Area of stationary object
If disparity estimation is performed on images with saturation, then complete scene coverage is achieved, but noise variations and estimation errors increase
Solution Approach 1:
The patent implements feedback by using estimated saturation levels to guide the disparity estimation process. The saturation information feeds back into the disparity calculation, allowing the system to adjust its processing strategy based on the reliability of each pixel region. This feedback mechanism ensures that disparity estimation is performed primarily on non-saturated, reliable pixels while still achieving complete scene coverage through multi-camera fusion.
Solution Approach 2:
The patent applies local quality by treating different regions of the image with different processing strategies based on their saturation characteristics. Non-saturated regions undergo full disparity estimation, while saturated regions are either excluded or processed with reduced confidence. This local differentiation maintains high reliability in good regions while still achieving comprehensive scene coverage through the combined output of multiple cameras.
Data Source
Figure 1
Figure 2A~2B
Figure 3A~3E
AI summary
Techniques for improved focusing of camera arrays are described. In one embodiment, a system may include a processor circuit, a camera array, and an imaging management module for execution on the processor circuit to capture an array of images from the camera array, the array of images comprising first and second images taken with first and second values of an exposure parameter, respectively, the first value different than the second value, to estimate a noise level, to normalize an intensity of each image based upon the noise level of the respective image, to produce a respective normalized image, to identify candidate disparities in each of the respective normalized images, to estimate a high dynamic range (HDR) image patch for each candidate disparity, and to compute an error from the HDR image patch and an objective function, to produce a disparity estimate. Other embodiments are described and claimed.