DCT Patch Descriptors for Real-Time Image Disparity Estimation
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing disparity estimation methods for images with shared micro-lenses are computationally complex and require software implementation, and are not adaptable to inputs with Bayer color patterns, leading to inefficiencies and artifacts in out-of-focus areas.
Innovation Solution
A hardware-efficient method using discrete cosine transform (DCT) coefficients to estimate disparity by generating a mean DCT descriptor from interleaved image patches, which are subdivided and analyzed using a classifier for real-time depth estimation.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If existing disparity estimation methods are used, then measurement precision is improved, but device complexity increases and productivity decreases
Solution Approach 1:
The image is divided into patches that are processed independently using DCT transforms. Each patch is analyzed separately to compute local disparity values, which are then combined to form the complete disparity map. This segmentation reduces computational complexity by breaking down the global optimization problem into smaller, more manageable local problems.
Solution Approach 2:
The patent replaces complex iterative optimization algorithms with a direct DCT-based approach. Instead of using heavy computational methods like dynamic programming or graph cuts, the invention uses discrete cosine transform coefficients to directly estimate disparity, significantly reducing computational requirements while maintaining acceptable accuracy.
2Measurement precision
If existing disparity estimation methods are used, then measurement precision is improved, but productivity decreases
Solution Approach 1:
By segmenting the image into patches and processing them independently with efficient DCT operations, the method achieves real-time processing speeds. The localized nature of patch-based processing allows for parallel computation, significantly improving productivity compared to global methods that must process the entire image sequentially.
Solution Approach 2:
The patent changes the computational parameters from iterative optimization steps to direct DCT coefficient analysis. This parameter change transforms the computational process from one requiring many iterations to one that can be completed in a fixed, small number of operations, enabling real-time disparity estimation.
3Device complexity
If local support is used for disparity estimation, then device complexity is reduced, but measurement precision worsens
Solution Approach 1:
The patent uses a small local support window for DCT analysis, which is sufficient for most regions. In areas where higher precision is needed, the method can combine results from multiple overlapping patches or apply additional refinement steps, achieving high accuracy without requiring large support windows throughout the entire image.
Data Source
AI summary
A method for disparity estimation between digital images includes generating an interleaved image from two or more digital images with an offset between them, where the interleaved image is subdivided into a plurality of patches, computing discrete cosine transform (DCT) coefficients of each of the plurality of patches, computing, for each of the plurality of patches, a mean DCT descriptor from the DCT coefficients of each patch, and determining a disparity map from the mean DCT descriptor of each of the plurality of patches using a classifier. The disparity map is configured for real-time depth estimation from the two or more digital images.