More Gemini Features Coming to Google TV
Google TV integrates Gemini AI for image/video generation, enhances Google Photos, and adds a dedicated YouTube Shorts feed to the homepage.
Read moreSo far, shooting outdoor scenes at night has always been something that can only be done with large cameras (such as DSLRs), although the night shooting technology of mobile phone cameras has made some progress, it is still not enough to satisfy people.
Google's Night Sight, which was launched last year, is part of the camera app of Google Pixel 3, which allows photographers to take beautiful pictures in dark environments. The normal camera mode produces grainy spots and severe underexposure. Google described how Night Sight achieves this in a blog post and discussed the technology at SIGGRAPH Asia 2019.
The amount of light detected by the image sensor of a mobile phone camera inherently has some uncertainty, called "grain noise," which makes the image appear grainy. The visibility of grain noise decreases as the amount of light increases; therefore, it is best for the camera to collect as much light as possible to produce high-quality images.
Within a given time, the amount of light reaching the image sensor is limited by the camera lens aperture. Extending the exposure time of the photo increases the total amount of light captured, but if the exposure time is too long, the motion in the shooting scene and the instability of the handheld camera will cause blurring. To overcome this problem, Night Sight divides the exposure into multiple frames, which have shorter exposure times and correspondingly less motion blur. The frames are first aligned to compensate for camera shake and motion within the scene, and then averaged and carefully processed for cases where perfect alignment is not possible. Although individual frames may be quite rough, the merged average image appears much sharper.
Shortly after the release of Night Sight, they started shooting stars in extremely dark outdoor environments. They realized that, just like previous experiments, high-quality images require several minutes of exposure time. Obviously, this is not suitable for mobile phone photography.
At the same time, night landscape photography must take into account scenes with motion. Viewers can tolerate blurry clouds and tree branches in photos, but blurry stars that look like short line segments look very bad. To mitigate this, they divide the exposure into several frames, with exposure times short enough to make the stars appear as points of light. When shooting night sky photos, they found that the exposure time for each frame should not exceed 16 seconds.
A 16-second exposure can capture enough light to generate recognizable images, but a night vision camera app must deal with other issues specific to low-light photography.
The phone uses its screen as an electronic viewfinder to capture a continuous stream of frames, which are displayed as real-time video to assist in lens synthesis. The camera's autofocus, auto exposure, and auto white balance systems use these frames simultaneously.
To aid in composition in extremely low light, Night Sight displays a "post-shutter viewfinder." After pressing the shutter button, each long-exposure frame is immediately displayed on the screen. These frames have exposure times of up to 16 seconds, capturing nearly 250 times more light than frames collected by a regular viewfinder, so after capturing the first frame, photographers can easily examine image details. They can then adjust the composition by moving the phone while the exposure continues. Once the composition is correct, the initial capture can be stopped and the second photo can be taken with all frames having the desired composition.
Autofocus ensures that the images captured by the camera are sharp. In normal operation, the incoming viewfinder frame is analyzed to determine how far the lens must be from the sensor to produce a focused image. However, in extremely weak light, the viewfinder frame may be too dark and grainy for detectable image details due to a lack of autofocus. In such cases, Night Sight on Pixel 4 switches to "post-shutter autofocus." After the user presses the shutter button, the camera captures two autofocus frames with exposure times of up to one second, which are sufficient to detect image details even in low light. These frames are used only for focusing the lens and do not directly contribute to the final image.
When viewing extremely dark environmental images on the screen, they appear much brighter than the original scene. This can change the viewer's perception of when the photo was taken during the day. At night, we expect the sky to be dark. If photos taken at night show a bright sky, we perceive them as scenes during the day, perhaps with slight abnormal illumination.
In Night Sight, this effect is counteracted by selectively darkening the sky in dimly lit scenes. To do this, machine learning is used to detect which areas of the image represent the sky. A convolutional neural network on the device has been trained to identify each pixel in the photo as "sky" or "non-sky" by tracking the contours of sky regions in 100,000 manually labeled images.
Sky detection also enables sky-specific noise reduction and selectively increases contrast to make features such as clouds, color gradients, or the Milky Way more prominent.
Of course, the capabilities of a mobile phone are not unlimited, and there is always room for improvement. Although night scenes are generally darker, they often contain bright light sources such as the moon, distant streetlights, or famous stars. Although we can capture landscapes under moonlight or details on the surface of the moon, the brightness range is very large (can exceed 500,000:1), which prevents us from capturing the same image simultaneously. Additionally, when stars are the only source of illumination, we can capture clear sky photos, but the landscape can only be seen as contours.
For Pixel 4, we have been using the brightest part of the Milky Way, near the constellation Sagittarius, as a benchmark for the quality of moonless sky images. By this standard, Night Sight performs well. Although the photos of the Milky Way show some residual noise, they are pleasing to the eye and show more stars and more details than what can be seen in the real night sky.
Google TV integrates Gemini AI for image/video generation, enhances Google Photos, and adds a dedicated YouTube Shorts feed to the homepage.
Read moreAlphabet Q1 paid subscribers reached 350M, fueled by YouTube & Google One. YouTube ad revenue slightly missed at $9.88B amid Premium growth. Total Q1 revenue hit $109.9B, beating forecasts.
Read moreAnthropic releases Claude connectors for Adobe, Blender, and Ableton, enabling AI workflow automation and app control. It also sponsors the Blender Development Foundation.
Read moreApple introduces an App Store model splitting 12-month plans into monthly installments. On iOS 26.4+, it lowers upfront costs without reducing total price, adding billing tracking.
Read moreGoogle launches Aadhaar digital IDs in Wallet for India, enabling secure QR/NFC verification. A phased partner rollout targets key sectors, while passport-based credentials expand globally.
Read moreYouTube tests "Ask YouTube," an AI tool blending text and video for complex queries. Initially for US Premium users, it supports interactive follow-ups, expanding Google's AI search.
Read moreGoogle grants US DoD classified AI access, joining OpenAI and xAI in defense deals amid Anthropic's dispute. Contractual safeguards against surveillance and lethal weapons remain unenforceable.
Read moreMistral AI launches Workflows, a Temporal-powered orchestration platform for secure, Python-driven enterprise AI automation featuring human-in-the-loop approvals.
Read moreCustom subliminal audio maker

AI Social Media Manager

AI Search Visibility

Workout and Calorie Tracker
Stock research platform

AI interview assistant

Scientific figure generator

AI courses with real code