Image Processing Device for Mobile Objects Using Color-Value Mean Correction
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing image processing devices for mobile objects that generate peripheral images by combining images from multiple imaging units face high calculation loads and insufficient image quality improvement, as they correct images based solely on luminance values.
Innovation Solution
An image processing device that corrects images using single-image color-value mean values from overlapping regions, reducing calculation load and unnatural color representations by setting reference values for lateral and rear-side images based on front-side images, and determining correction thresholds to prevent erroneous adjustments.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Manufacturing precision
If image correction is performed based on luminance values of entire images, then image quality improvement is attempted, but calculation load becomes large and insufficient correction results are obtained
Solution Approach 1:
The patent divides the image correction process into two stages: first correcting luminance values based on overlapping regions, then correcting color values based on target regions. This segmentation reduces the calculation load compared to correcting entire images while maintaining image quality improvement.
Solution Approach 2:
The patent applies different correction strategies to different parts of the image: luminance correction is applied to overlapping regions where images meet, while color correction is applied to target regions representing road surfaces. This local quality approach optimizes correction effectiveness while reducing unnecessary calculations.
2Manufacturing precision
If correction is applied to all images including front-side images, then color consistency is improved, but visual unnaturalness increases for the driver
Solution Approach 1:
Instead of correcting all images to match a reference, the patent inverts the approach by using the front-side image (which the driver directly views) as the reference and only correcting lateral and rear-side images to match it. This prevents visual unnaturalness while maintaining color consistency.
Solution Approach 2:
The patent copies the color characteristics of the front-side image to the lateral and rear-side images through correction, rather than forcing all images to conform to a theoretical standard. This preserves the natural appearance of the driver's direct view while achieving color consistency across the peripheral image.
3Measurement precision
If correction values are calculated without thresholds, then correction accuracy is maximized, but erroneous corrections occur on road surfaces and white lines
Solution Approach 1:
The patent introduces threshold parameters (color-value mean value threshold and variation threshold) to control the correction process. These parameters dynamically adjust whether correction is applied based on the characteristics of the target region, preventing erroneous corrections on road surfaces and white lines while maintaining accuracy in appropriate regions.
Solution Approach 2:
The patent uses feedback mechanisms by calculating color-value mean values and variations, then comparing them against thresholds to determine whether correction should be applied. This feedback loop ensures that correction is only performed when appropriate, improving reliability while maintaining accuracy.
Data Source
AI summary
According to one embodiment, the image processing device includes imagers disposed on an outer circumference of a mobile object for imaging surroundings of the mobile object to generate multiple images including mutually overlapping regions, and a processor that corrects the images on the basis of one color-value mean value and another color-value mean value to generate a peripheral image by combining the corrected images. The one color-value mean value is an average of color values of a target region set in an overlapping region of one of the images. Another color-value mean value is an average of color values of a target region set in an overlapping region of another one of the images.