Stereo Depth Estimation With Asymmetric Downsampling for Disparity
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing stereo depth estimation methods suffer from reduced accuracy due to equal down-sampling in height and width, leading to loss of critical details and computational inefficiency.
Innovation Solution
Implement asymmetric down-sampling operations, prioritizing higher resolution in width over height, to preserve disparity and enhance resolution while maintaining efficiency.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Device complexity
If equal down-sampling is performed in height and width, then computational complexity is reduced, but depth estimation accuracy deteriorates due to loss of critical details
Solution Approach 1:
The patent applies asymmetric down-sampling where the down-sampling ratio differs between height and width dimensions. Specifically, it uses a larger down-sampling ratio in the height direction and a smaller down-sampling ratio in the width direction, breaking the symmetry of conventional equal down-sampling. This asymmetry preserves more horizontal details critical for disparity estimation while still achieving computational reduction through vertical down-sampling.
Solution Approach 2:
The patent recognizes that different spatial dimensions have different importance for stereo depth estimation. The width dimension contains more critical disparity information, so it is preserved with higher resolution (smaller down-sampling ratio). The height dimension can be down-sampled more aggressively without losing essential depth estimation accuracy. This local quality differentiation optimizes the balance between computational efficiency and accuracy.
2Adaptability or versatility
If down-sampling is performed at multiple pyramid levels, then feature extraction is enhanced, but critical details are lost leading to reduced model accuracy
Solution Approach 1:
The patent extends asymmetric down-sampling to multiple pyramid levels, where each level uses different down-sampling ratios for height and width. This maintains the asymmetric preservation of width-direction details across all pyramid levels, ensuring that critical disparity information is retained even as feature extraction capability is enhanced through multi-level processing.
3Productivity
If resolution is reduced to lower computational complexity, then processing efficiency is improved, but accuracy and disparity estimation hypotheses are significantly reduced
Solution Approach 1:
The patent implements asymmetric down-sampling that reduces overall resolution and computational complexity while preserving the width dimension resolution more than the height dimension. This asymmetric approach ensures that disparity estimation hypotheses remain sufficiently accurate in the critical width direction where most disparity information resides, while still achieving processing efficiency gains through height direction down-sampling.
Data Source
AI summary
Aspects relate to stereo depth estimation utilizing asymmetric down-sampling in different directions. A device may include one or more memories configured to store a plurality of images and a plurality of cameras. The plurality of cameras may be configured to capture a left and right image, in which, each of the images includes one or more patches, each patch including plurality of pixels. The device may include one or more processors coupled to one or more memories, in which, the one or more processors are configured to: down-sample in a first direction on a first set of pixels in a first patch of a first image to generate a first down-sample; and down-sample in a second direction on a second set of pixels in a second patch of a second image to generate a second down-sample, the second down-sample including a greater number of pixels.