Sony Patents an AI That Fuses Invisible Light Into Your Photos to Fix Them
Your camera can only see so much. Sony is patenting a system that uses invisible wavelengths of light, captured alongside a normal photo, to fill in what the visible-light sensor missed.
What Sony's multispectral photo fix actually does for your pictures
Every time your phone snaps a photo in bad lighting, it's working with whatever light happens to be bouncing off the scene at that moment. If the room is dim, the photo comes out grainy. If the subject is backlit, the shadows lose detail. The camera is stuck with what it can see.
Sony's patent describes a chip and software setup that pairs your ordinary visible-light photo with a second image captured in wavelengths your eyes can't see, like near-infrared. A trained AI model then reads both images together and produces a cleaned-up version of the visible photo, with better sharpness, less noise, or improved color, depending on what it finds useful in the invisible-light data.
Think of it as giving the AI two windows into the same scene. The regular photo tells it what colors look like. The multispectral image tells it where edges, textures, and depth really are. The result is a single photo that looks better than either sensor could produce alone.
receive first image data representing a multispectral image of a scene; and receive second image data representing a first image of the scene in the visible light spectrum; …
Translation: The system takes in invisible light data along with a normal photo.
How the model blends multispectral and visible-light frames
The patent describes a hardware-plus-software system with two main components: interface circuitry (the part that receives image data) and processing circuitry (the part that runs the AI).
The interface takes in two streams at once:
- A multispectral image, a capture that records light across multiple wavelengths, including ones outside the visible range, such as near-infrared or ultraviolet. These wavelengths can reveal surface texture, material properties, and structural edges that a normal camera sensor ignores.
- A visible-light image, the ordinary photo you'd expect a camera to produce, capturing the colors and tones the human eye sees.
The processing circuitry feeds both images into a trained machine-learning model. That model, having already learned from a large set of image pairs during a separate training phase, uses the multispectral data to guide improvements to the visible-light photo. The output is a second visible-light image with at least one "enhanced image property", the patent's language is intentionally broad, covering sharpness, noise reduction, color accuracy, or dynamic range.
A companion part of the patent covers how such a model is trained, meaning Sony is claiming both the inference engine and the training pipeline that builds it.
… generate, based on the multispectral image of the scene and the first image of the scene in the visible light spectrum and using a trained machine learning model for image enhancement, a second image of the scene in the visible light spectrum with at least one enhanced image property …
Translation: An AI uses that invisible data to output a much better normal photograph.
What this means for cameras in phones, cars, and security gear
For most people, this would show up as a camera that handles dark rooms, harsh outdoor contrast, or fast-moving subjects better than today's phones or security cameras can. The improvement wouldn't come from a bigger sensor or a brighter flash, it would come from the AI reading light your eyes can't see and using it to repair the photo you can.
Sony's steady investment in image sensor AI points toward this kind of on-chip processing becoming a selling point for the image sensors Sony supplies to smartphone makers, automotive camera systems, and surveillance hardware. The practical payoff for you is photos that need less editing and cameras that work in conditions where current hardware just gives up.
This is the 12th Sony filing we've tracked since July in the AI photo editing race, following one on camera settings from your feedback and one on rebuilding 3D scenes from video.
If your indoor photos keep coming out dark, blurry, or washed out, this is Sony working on the underlying reason why: cameras simply do not have enough information to work with in dim conditions. The idea here is to pull in light the human eye cannot see, use it to fill in what the visible image is missing, and hand you a better photo automatically.
You would never interact with any of this directly. A future camera using this approach would just produce sharper, cleaner shots in the kinds of rooms where photos usually disappoint.
The honest caveat is that a patent describes an intention, not a finished result. The moment this actually matters to a real person is when they pull out their phone at a dinner table and the photo looks the way they hoped it would.
There are more where this came from
We read every patent application Big Tech publishes and send you the ones worth knowing. Plain English, free, every week.
The drawings
6 drawing sheets from US 2026/0268452 A1 · click any drawing to enlarge
Want this weekly breakdown for a company we don't cover? Patentlyze Pro →