Harmonized Combined Image Correction via Pixel Statistics
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing systems for combining images from multiple cameras on a vehicle require significant processing resources to adjust for differences in brightness and color, leading to potential delays in displaying a harmonized 360° view.
Innovation Solution
A method that calculates a correction factor based on pixel statistics in overlapping regions of adjacent camera images, applying it in hardware devices to reduce differences in pixel statistics by less than half, thereby minimizing processing requirements and avoiding delays in image display.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Manufacturing precision
If correction gain values are calculated for each image from multiple cameras to harmonize brightness and color, then image harmonization quality is improved, but processor resource consumption increases
Solution Approach 1:
The system calculates correction gain values in advance based on representative pixels from each camera, storing these values for subsequent image processing. This preliminary calculation approach allows rapid application of harmonization corrections to multiple images without recalculating gain values each time, significantly reducing processor resource consumption during actual image display operations.
Solution Approach 2:
Instead of processing all pixels in each camera image, the system selects and processes only representative pixels (sampling pixels) from each image to calculate correction gain values. These correction values are then applied to the entire image or image regions, effectively copying the correction from sampled points to the full image data, thereby dramatically reducing computational load while maintaining harmonization quality.
2Manufacturing precision
If correction gain values are calculated for each image to harmonize adjacent camera images, then image uniformity is improved, but processing time increases
Solution Approach 1:
The system performs correction gain calculation in advance during periods when image processing resources are available, storing the calculated gains for rapid application. This separates the computationally intensive calculation phase from the time-critical display phase, ensuring that actual image harmonization during driving does not cause delays.
Solution Approach 2:
The system calculates correction gains based on representative pixels and copies these correction values across the entire image or overlapping regions. This approach maintains image uniformity while minimizing processing time by avoiding pixel-by-pixel calculation during critical display intervals.
3Speed
If multiple cameras capture images at high frequency to provide continuous 360° view, then view continuity is improved, but processor resource requirements increase
Solution Approach 1:
The system processes only representative pixels from each high-frequency camera image to calculate or update correction gain values, then applies these corrections to the full images. This sampling approach enables continuous 360° view processing at high frequencies without proportionally increasing processor resource requirements.
Solution Approach 2:
The system maintains a cache of previously calculated correction gain values that can be applied to subsequent images without recalculation. This allows the system to handle high-frequency image streams from multiple cameras by reusing correction data, significantly reducing the processing burden of continuous 360° view generation.
Data Source
AI summary
A method for harmonizing a combined image involves receiving a respective image or image frame from each of two or more cameras, the images from two of the cameras representing the same region in an overlap region, measuring pixel statistics of at least some of the pixels in the overlap region for the image of each of the two cameras, determining a difference in the pixel statistics Δpixel of the image from each of the two cameras in the overlap region, calculating a correction factor Kpixel, wherein the correction factor Kpixel may be predicted to produce a reduction in Δpixel of less than Δpixel/2 when applied to the image of one of the two cameras, applying the correction factor Kpixel in a hardware device acquiring the image, and receiving a further image or image frame from each of the two cameras.