Recently, members of the Google Research team presented RawNeRF, an AI-based camera software that could improve smartphone computational photography. Introduced during the Computer Vision and Pattern Recognition Conference (CVPR), RawNeRF is designed to work with smartphone camera sensors, which are typically much smaller than those found in DSLR cameras. As a result, manufacturers have looked towards computational photography for image quality improvements, particularly in low-light conditions.
RawNeRF expands upon this method by working with several RAW images taken from different angles. From there, it calculates almost noise-free HDR night shots. Google emphasises that RawNeRF is not just a denoiser, either. Supposedly, RawNeRF results can be customised, such as exposure and gamma levels, focus and even perspective. Additionally, the software can reconstruct missing information, unlike other AI photography solutions.
Unfortunately, it could be a while yet before Google brings RawNeRF to the masses. Currently, the software is in the research stage of development. Hence, it is unlikely that Google will introduce the software with the upcoming Pixel 7 series or even next year's Pixel 8 series. However, we would not be surprised if Google integrated RawNeRF somehow in future Pixel smartphones, given its preference to utilise older or lower resolution camera sensors for multiple generations while perfecting its camera software.
Source(s)
GitHub via 9to5Google