HDR Image Generation via Continuous Weighting Maps
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Standard image sensors have a limited dynamic range, leading to information loss when capturing scenes with luminance beyond their range, resulting in poor image quality due to saturation or underexposure, and existing HDR technologies often introduce artifacts like discontinuities in smooth transitions.
Innovation Solution
A method and system for generating high dynamic range images by determining pixel values as a weighted average of input images from different brightness levels, using a weighting map to assign continuous combination weights based on pixel intensities, scaling, and normalization, which avoids discontinuities in smooth transitions.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Illumination intensity
If multiple exposures are merged to increase dynamic range, then the captured dynamic range is improved, but discontinuities and artifacts appear in smooth transition regions
Solution Approach 1:
The patent introduces a weighting map as an intermediary structure that continuously assigns weights to different input images based on pixel intensity values. This weighting map acts as a mediator between the multiple exposures, smoothly transitioning between them without abrupt changes, thereby eliminating discontinuities while preserving the expanded dynamic range.
Solution Approach 2:
The patent changes the parameter of combination weights from discrete to continuous as a function of pixel intensity. By making the weights continuous rather than discrete, the transition between different exposure images becomes smooth, preventing artifacts while maintaining the benefits of multiple exposure merging for extended dynamic range.
2Device complexity
If discrete weighting is used to combine images, then processing simplicity is improved, but discontinuities appear in smooth transitions
Solution Approach 1:
The patent transforms the weighting parameter from discrete to continuous by defining combination weights as continuous functions of pixel intensity values. This parameter change enables smooth transitions between different input images while maintaining computational feasibility through the use of predefined weighting maps.
3Ease of manufacture
If standard sensors are used to capture scenes, then cost and affordability are improved, but information is lost in bright or dim regions
Solution Approach 1:
The patent segments the luminance range by capturing multiple exposures at different brightness levels. Each exposure is optimized for a specific portion of the luminance range, allowing standard sensors to capture information across the full dynamic range of the scene by dividing the capture task across multiple images that are then merged.
Solution Approach 2:
The patent performs preliminary capture actions by taking multiple exposures before merging them. Each exposure is预先 captured at different brightness levels to ensure that all regions of the scene (both bright and dim) are properly exposed in at least one image, preventing information loss before the combination process.
Data Source
AI summary
A method determines a pixel value in a high dynamic range image from two images of different brightness by obtaining corresponding input pixel intensities from the two images, determining combination weights, and calculating the pixel value in the high dynamic range image as a weighted average of the input pixel intensities. Another method determines a pixel value in a high dynamic range image from more than two images by forming pairs of corresponding input pixel intensities, determining relative combination weights for the input pixels intensities for each pair, applying a normalization condition to determine absolute combination weights, and calculating the pixel value in the high dynamic range image as a weighted average of the input pixel intensities. Systems for generating high dynamic range image generation from two or more input images include a processor, a memory, a combination weight module, and a pixel value calculation module.