Clear Pixel Noise Correction for Lane Recognition
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing image processing techniques for lane partition line recognition in vehicle cameras suffer from noise issues, particularly at night, due to increased digital gain, leading to reduced contrast and increased processing load, as seen in interpolation methods like those in Japanese Patent Applications Laid-Open Publication No. 2009-55281 and No. 2009-206553.
Innovation Solution
An image processing apparatus that acquires image data with red, blue, green, and clear pixels, determines if the luminance value of a pixel of interest is abnormal by comparing it to the closest clear pixel's value, and corrects it if necessary, reducing noise and processing load by avoiding the clouding of pixel colors and minimizing the sorting of luminance values around the pixel of interest.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If interpolation process is performed using pixels away from pixel of interest to reduce noise, then noise is reduced, but color of pixel of interest is clouded and contrast is lowered
Solution Approach 1:
The patent applies local quality by using only clear pixels (C-pixels) that are spatially adjacent to the pixel of interest for noise correction, rather than using pixels from a broader area. This localized approach ensures that the correction is applied only where necessary and maintains the color integrity of the pixel of interest while reducing noise. The clear pixels serve as reliable reference points that do not introduce color clouding.
2Reliability
If interpolation process is performed by sorting luminance values of 25 pixels around pixel of interest to reduce noise, then noise is reduced, but processing load is increased
Solution Approach 1:
The patent extracts only the necessary information for noise reduction by using luminance values from clear pixels (C-pixels) that are spatially adjacent to the pixel of interest. Instead of processing all 25 surrounding pixels as in conventional methods, the invention selectively uses only the clear pixels, which are fewer in number and easier to process. This extraction approach significantly reduces the processing load while maintaining effective noise reduction.
3Illumination intensity
If digital gain is increased during night hours to improve image brightness, then image brightness is improved, but noise spikes are accentuated
Solution Approach 1:
The patent introduces clear pixels (C-pixels) as intermediaries to mediate between the noisy colored pixels (R, G, B pixels) and the final image output. The C-pixels, which have lower noise characteristics, serve as reference points to correct the luminance values of the noisy pixels. This intermediary approach allows the system to maintain the brightness enhancement from digital gain while using the cleaner C-pixel data to suppress noise spikes in the final image.
Data Source
AI summary
In an image processing apparatus, an image data acquisition unit acquires image data including R-pixels which are red pixels, holding red luminance values, B-pixels which are blue pixels, holding blue luminance values, G-pixels which are green pixels, holding green luminance values, and C-pixels which are clear pixels, holding luminance values in a wavelength range including wavelength ranges corresponding to red, blue and green. In the apparatus, a determination unit determines whether or not an inequality given by Valc−Valx<α is satisfied, where Valx is a luminance value of a pixel of interest that is one of the R-, B-, and G-pixels in the image data, Valc is a luminance value of the C-pixel or pixels closest to the pixel of interest, and α is a constant, and if the inequality is satisfied, a correction unit corrects the luminance value Valx to be less than the luminance value Valc.