Image Processing Apparatus Log-Scale Demosaicking
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing image processing methods degrade image quality due to artifacts like false colors and zipper noise when demosaicking linear-scale raw signals, as color differences vary with luminance, leading to poor image quality, especially when gamma correction is performed after color gamut conversion.
Innovation Solution
An image processing apparatus that performs gamma correction on linear-scale raw signals to generate log-scale signals, demosaics these log-scale signals, and then applies inverse gamma correction to produce high-quality image signals, thereby reducing artifacts and maintaining image quality during color gamut conversion.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Productivity
If demosaicking is performed on linear-scale raw signals, then the processing can be completed in the conventional order, but artifacts such as false colors and zipper noise occur due to varying color differences with luminance
Solution Approach 1:
The patent applies preliminary action by performing gamma correction on the raw signals before demosaicking. This preliminary transformation converts the linear-scale signals to log-scale signals, which stabilizes the color differences across different luminance levels. By preparing the signals in advance through gamma correction, the subsequent demosaicking process can proceed without generating artifacts, thus resolving the contradiction between completing processing and maintaining image quality.
Solution Approach 2:
The patent employs parameter changes by transforming the signal scale from linear to log through gamma correction. This change in the mathematical parameter (signal representation) fundamentally alters how color differences are perceived and processed. By changing the parameter space, the patent enables stable color difference representation across luminance ranges, eliminating artifacts during demosaicking while maintaining processing completeness.
2Ease of operation
If gamma correction is performed after color gamut conversion, then the conventional processing order is maintained, but image quality degrades due to the need to perform color gamut conversion in linear scale
Solution Approach 1:
The patent applies preliminary action by performing gamma correction before demosaicking and color gamut conversion. This preliminary log-scale transformation enables subsequent color gamut conversion to be performed in a stable color space where color differences are luminance-independent. By preparing the signal state in advance, the patent allows color gamut conversion to be performed after gamma correction without degrading image quality, thus resolving the contradiction between maintaining processing order and preserving image quality.
3Adaptability or versatility
If color gamut conversion is performed in linear scale, then the conversion can be applied to raw signals, but artifacts occur because color differences vary with luminance
Solution Approach 1:
The patent employs parameter changes by transforming signals from linear scale to log scale through gamma correction before color gamut conversion. This parameter transformation fundamentally changes the signal characteristics, making color differences independent of luminance. By changing the mathematical representation, the patent enables color gamut conversion to be applied while maintaining color accuracy across different luminance levels, thus resolving the contradiction between conversion applicability and color accuracy.
Data Source
AI summary
An image processing apparatus includes a gamma correction unit configured to perform gamma correction on linear-scale raw image signals to generate log-scale signals, a demosaicing unit configured to demosaic the log-scale signals generated by the gamma correction unit to generate image signals, and an inverse gamma correction unit configured to perform inverse gamma correction on the image signals generated by the demosaicing unit to generate linear-scale signals.