Confidence Map Depth Boundary Upsampling
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing methods for up-sampling depth and motion images to match the spatial resolution of photographed images often result in blurred depth boundaries due to incorrect averaging of pixels with similar colors, and increase calculation volume through edge extraction and filter processing, leading to reduced accuracy.
Innovation Solution
An image data processing apparatus and method that determines the level of confidence in each pixel of expanded image data using a comparison between first and second evaluation values based on neighboring pixels, with different methods for determining these values, and corrects depth information using a confidence map to exclude unreliable pixels, thereby reducing calculation volume and improving precision.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If up-sampling is performed to increase spatial resolution, then the amount of information increases to match the photographed image, but information on the boundary portion of depth or motion cannot be accurately acquired
Solution Approach 1:
The patent segments the image into multiple regions based on depth boundaries and processes each region differently. By identifying boundary portions through confidence map analysis, the system applies selective up-sampling strategies: preserving original boundary information while applying interpolation only to non-boundary regions, thus maintaining boundary accuracy while achieving high spatial resolution elsewhere.
Solution Approach 2:
The patent implements local quality by applying different processing methods to different spatial locations. Confidence maps are generated to identify reliable versus unreliable pixel regions, and up-sampling is selectively applied only to regions with sufficient confidence. This ensures that boundary portions maintain their original accuracy while other regions achieve enhanced resolution.
2Measurement precision
If joint bilateral filter is used to correct depth information during up-sampling, then depth boundary accuracy is improved, but calculation volume increases
Solution Approach 1:
The patent applies partial action by using a simplified filtering approach only where necessary. Instead of applying joint bilateral filter to the entire image, the system generates confidence maps to identify specific boundary regions and applies correction only to those areas. This reduces the overall calculation volume while maintaining depth boundary accuracy in critical regions.
Solution Approach 2:
The patent extracts boundary information from the depth map using confidence map analysis, separating boundary regions from non-boundary regions. By extracting only the essential boundary information and processing it separately with simplified methods, the system avoids the heavy computational burden of applying complex filters to the entire image while still achieving accurate depth boundary representation.
3Reliability
If Gaussian filter and confidence map generation are used to prevent averaging errors, then boundary accuracy is maintained, but calculation volume and processing complexity increase
Solution Approach 1:
The patent implements self-service by using the depth map itself to generate the confidence map through simple statistical analysis of neighboring pixels. Instead of requiring complex external edge detection algorithms or multiple filtering passes, the system derives confidence information directly from the depth data by comparing pixel values with their neighbors, thereby maintaining boundary accuracy with minimal additional processing complexity.
Solution Approach 2:
The patent changes the parameter representation by transforming depth values into confidence values through a straightforward computational process. By calculating confidence based on the variance or difference between neighboring pixel values, the system creates a confidence map that automatically highlights boundary regions without requiring complex filtering operations, thus maintaining accuracy while reducing processing complexity.
Data Source
AI summary
An image data processing apparatus includes: an image data acquisition unit for acquiring low resolution image data and high resolution image data; and a confidence data determination unit for determining confidence data which indicates a level of confidence in a pixel value in the high resolution image data. The determination unit determines the level of confidence of a target pixel in the high resolution image data based on a comparison between a first and a second evaluation value. The first evaluation value is determined based on pixel values of neighboring pixels in a pixel position in the low resolution image data corresponding to a pixel position of the target pixel in the high resolution image data. The second evaluation value is determined based on the pixel values of the neighboring pixels, using a method that is different from the first evaluation value.