Why does the camera keep the G component unchanged during white balance adjustment?
The design of fixing one component when adjusting RGB during white balance processing primarily stems from the following four reasons:
(1) Non-uniqueness of RGB parameter combinations
The white balance color is determined by the ratio of the three RGB components and is not directly related to their actual values. Therefore, there can be countless combinations that achieve the same white balance—same color but different brightness levels. For example, three colors with RGB values of (100, 100, 100), (128, 128, 128), and (255, 255, 255) all have a ratio of 1:1:1, meaning they share the same color but differ in brightness. If the values are too small, the overall image may appear too dark; if the values are too large, the image may become overexposed.
(2) Algorithm optimization
In conventional algorithms, fixing one component improves the efficiency of the ISP algorithm, prevents data overflow, avoids color abnormalities, and enhances system reliability. Fixing one component reduces the degrees of freedom, simplifying the three-dimensional color space adjustment to a two-dimensional mapping problem. For instance, by fixing the G component at a reasonable value, white balance can be achieved by adjusting the gains of R and B. For example, when the white balance color is biased toward green (i.e., too much green and too little red and blue), increasing the R and B components can achieve the desired white balance effect. This design ensures theoretical rationality while significantly reducing computational complexity.
(3) Sensor Physical Structure Characteristics
Mainstream sensors employ a Bayer array color filter layout (RGGB arrangement), where the number of G pixels is twice that of R or B. This design grants the G channel a higher sampling density, capturing over 60% of the overall luminance information. As a result, it establishes a more stable statistical baseline.
(4) Characteristics
The human retina is most sensitive to the green wavelength around 550nm, with approximately 70% of the visual system's luminance perception derived from G-channel information. Maintaining the G component as a benchmark preserves luminance features critical to human visual sensitivity, thereby avoiding noticeable luminance distortion during adjustment processes.
In summary, keeping the G component unchanged during white balance adjustment represents a trade-off between application requirements and algorithmic efficiency, rather than being a technical limitation. The method of fixing the G component remains the foundational implementation scheme for most ISPs today.
