Image Processing Method for CMOS Sensor Quality Improvement
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
The image quality generated by CMOS sensors with multi-pixels-in-one technology is generally low, requiring an improvement technique to enhance image quality.
Innovation Solution
An image processing method that includes receiving a raw image, performing remosaicing to convert it into a full resolution image in Bayer format, involving false color correction, high-frequency detail extraction, and post-processing based on illuminance conditions, using techniques like direction interpolation and upscaling of binning pixel images.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Adaptability or versatility
If multi pixels-in-one technology is used in CMOS sensor, then pixel integration is improved, but image quality deteriorates
Solution Approach 1:
The patent segments the image processing into distinct stages: raw image reception, remosaicing to generate Bayer format image, false color correction, and full resolution image generation. This segmentation allows each stage to be optimized independently, resolving the contradiction between pixel integration and image quality.
Solution Approach 2:
The patent performs preliminary remosaicing and false color correction on the raw image before generating the final full resolution image. This preliminary action prepares the image data in advance, enabling quality improvement without compromising the pixel integration capability.
2Measurement precision
If false color correction is performed on first color image, then color accuracy is improved, but over-correction occurs
Solution Approach 1:
The patent applies false color correction to the first color image and then performs additional correction on the resulting second color image to compensate for over-correction. This feedback mechanism ensures color accuracy while preventing distortion.
Solution Approach 2:
The patent deliberately applies excessive false color correction and then compensates for the over-correction in a subsequent step. This partial action approach ensures that color accuracy is achieved without permanent distortion.
3Manufacturing precision
If high-frequency details are extracted from second image, then image detail is improved, but processing complexity increases
Solution Approach 1:
The patent extracts high-frequency details from the second image before generating the final full resolution image. This preliminary extraction prepares the detail information in advance, reducing the complexity of the final image generation process.
4Manufacturing precision
If binning pixel image is upscaled to match resolution, then resolution compatibility is improved, but processing time increases
Solution Approach 1:
The patent performs upscaling of the binning pixel image to match the resolution of the high-frequency detail image before combining them. This preliminary upscaling ensures resolution compatibility is achieved in advance, streamlining the final image generation process.
Data Source
AI summary
An image processing method includes receiving a raw image of an image sensor as a first image; performing remosaicing on the first image to generate a second image in a Bayer format, the performing the remosaicing including: converting the first image into a first color image, performing false color correction on the first color image, re-correcting an over-correction caused in performing the false color correction to generate a second color image, and converting the second color image into the second image in the Bayer format; and generating a full resolution image as an output image based on the second image.