Local Contrast Enhancement via 2D-Separable SAD Approximation
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current image enhancement methods, particularly for real-time video applications, face challenges in efficiently calculating local contrast differences due to high computational complexity, which hinders the visibility of details in complex scenes and low-quality video footage, and often result in artifacts like ringing.
Innovation Solution
A method for calculating local energy measurements 2D-separably, approximating the sum of absolute differences (SAD) metric, and using a modified contrast gain function to prevent signal clipping and suppress artifacts, enabling perceptually uniform contrast enhancement.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Manufacturing precision
If locally adaptive contrast enhancement methods are used to improve image quality and visibility of details, then the perceived output image quality is improved, but the computational complexity becomes very high
Solution Approach 1:
The patent divides the image processing task into multiple scales using a pyramid structure, where the image is decomposed into coarse-to-fine levels. This segmentation allows contrast enhancement to be applied at different resolutions, reducing the computational burden at each level while maintaining overall image quality.
Solution Approach 2:
The patent implements locally adaptive contrast enhancement by computing local standard deviation values for different regions of the image and applying scale-specific gain factors. This allows different parts of the image to be enhanced with appropriate local contrast adjustment, improving detail visibility while managing computational complexity through selective processing.
2Manufacturing precision
If large filter kernels are used to achieve desired quality improvement in high-resolution images, then the image quality is improved, but the processing delay increases
Solution Approach 1:
The patent applies a multi-scale pyramid approach where large filter kernels are used only at coarse levels of the pyramid, while finer levels use smaller kernels. This segmentation of processing across scales reduces the overall processing delay compared to applying large kernels uniformly across the entire high-resolution image.
Solution Approach 2:
The patent performs contrast enhancement operations at coarser pyramid levels before processing finer levels. This preliminary action at reduced resolutions prepares the image data, allowing subsequent fine-level processing to be more efficient and reducing total processing time while maintaining quality.
3Manufacturing precision
If multi-scale approaches with filters of different scales are used to enhance details in different areas, then the detail enhancement is improved, but the device complexity increases
Solution Approach 1:
The patent organizes multi-scale processing in a structured pyramid hierarchy with defined levels and transition rules. This segmentation provides a systematic framework that manages the complexity of applying multiple filter scales, making the processing pipeline more organized and implementable compared to unstructured multi-scale approaches.
Data Source
Figure 1A~1B
Figure 1C~2
Figure 3A~3B
AI summary
A method for obtaining and/or enhancing local contrast differences (LD) in digital image input data (in), wherein a local contrast measure (SADap) approximating a two- dimensional sum of absolute difference (SAD) measure is obtained by obtaining a first- dimension sum of absolute difference measure (energyVk(m, n)) representing a sum of absolute difference (SAD) in a first dimension of at least one image region, and obtaining the local contrast measure (SADap) as a second-dimension convoluted first-dimension sum of absolute difference measure by convo luting the first-dimension sum of absolute difference measure (energyVk(m, n)) in a second dimension of the at least one image region, is proposed, wherein the step of obtaining the local contrast measure (SADap) as the second-dimension convoluted first-dimension sum of absolute differences measure further includes an addition step wherein an energy term (?lpfk(j)) representing a first- order difference on a distance (J) of a low pass filtered signal used in obtaining the first- dimension sum of absolute differences measure (energyVk(m, n)) and in obtaining the second-dimension convoluted first-dimension sum of absolute differences measure is introduced while obtaining the local contrast measure (SADap).