It is correct that purple fringing occurs when green (G) pixels in high-brightness areas of the sensor become saturated and contaminate unsaturated red and blue pixels, creating a purple tint at red and blue pixel positions. However, the same sensor paired with different lenses may show purple fringing of varying severity, or none at all. Therefore, purple fringing cannot be considered unrelated to the lens.

Apart from purple fringing, another common artifact is purple flare.

The root causes of purple flare are well defined: internal reflections inside the lens or camera module, stray light scattering, scattering at the aperture edge, lens surface roughness, and dust within the module. These factors cause high-angle incident light to activate not only the photodiodes (PD) at the intended pixel position but also neighboring photodiodes, eventually forming a colored halo, as illustrated below.

This phenomenon is known as optical crosstalk. Why do most chromatic flares appear purple? For an RGGB color filter array (CFA), the green color filter has higher transmittance than the red and blue filters. As a result, high-angle incident light passes through the green filter more easily and is absorbed by adjacent red and blue pixels. This raises the R/G and B/G ratios at these positions relative to normal pixels. After white balance and color correction matrix (CCM) processing, purple or pink halos appear in the final image.

For automotive and machine vision applications using sensors with alternative CFAs such as RCCB or RYYCy, chromatic flare occurs more frequently. This is because the C or Y filters have higher transmittance than conventional G filters.

Mitigation of chromatic flare involves both optical improvements to lenses and modules, as well as sensor pixel design optimizations to suppress optical crosstalk.

One method is adding metal separation grids to block crosstalk from high-angle light, as shown in the diagram below.

Another approach is deep trench isolation (DTI), where isolation structures are placed deeper within the pixel to block light crosstalk. The diagram below depicts backside deep trench isolation (BDTI).

How can engineers evaluate the crosstalk resistance of different sensors during sensor selection?

Sensor spectral characterization can display quantum efficiency (QE) response curves for each color channel under specific incident angles, as shown below.

The graph compares the RGB QE curves of an RGGB sensor under 30° incident light: the baseline device, the sensor with composite grid (CG) technology, and the sensor with both CG and BDTI technologies.

It is clear that the latter exhibits better separation between the R, G and B QE responses, proving effective crosstalk reduction.