CamCyclopedia
Contents
- 200MP Adaptive Pixel – Bright and clear pictures anytime anywhere
2. Multi-frame Image Processing with Neural Network – Overcoming the physical limitations of sensors
3. Galaxy S23 camera’s high resolution
4. Link with Camera Assistant
200MP Adaptive Pixel – Bright and clear pictures anytime anywhere
We have introduced a 200 million pixels (200MP) image sensor with ultra-high resolution technology in Galaxy S23 Ultra.
The increase in the number of pixels, in the same optical format, to 200 million has caused the light-reception area per pixel to decrease. It has increased the noise levels.
We have applied a '200MP Adaptive Pixel' technology to overcome these physical limitations and capture bright and sharp images anytime anywhere.
The 200MP Adaptive Pixel technology groups 4 or 16 neighboring pixels together to increase the light reception area, depending on lighting conditions. It switches to 50MP in High Dynamic Range (HDR) environments, such as mid-light or backlight. On the other hand, 12MP in low-light environments, such as at night, ensures there is enough light for optimal image quality.
As a result, you can take night-time photos with less noise in low-light conditions, but they may have less detail than results taken under bright conditions. If you select the 200MP mode, the image taken with the Adaptive Pixel technology image is up-scaled (50MP, 12MP) and saved as a 200MP. (The same technology for upscaling in low-light conditions is also available in the S22 108MP mode).
※. We're looking into adding an Adaptive Pixel On/Off option to Camera Assistant, allowing users to shoot at their preferred resolution.
Multi-frame Image Processing with Neural Network – Overcoming the physical limitations of sensors
Multi-frame image processing technology uses multiple frames to overcome the physical limitations of various sensors. Since noise is inevitably accompanied by small pixels, especially in high-resolution sensors, it is developed to remove noise and restore a limited dynamic range. This technology is a part of the Galaxy S23 series.
It's basically an algorithm that estimates and compensates for motion between frames, and then, on that basis, combines multiple frames. It reduces noise by taking multiple shots of a scene with the same exposure and stacking them together. It is effective in low-light situations.
The multi-frame denoising algorithm has been expanded as a technique for HDR restoration. It utilizes multiple images with different exposures, consisting of long and short exposures, to restore limited dynamic range. It also uses tone mapping, the process of mapping HDR images to typical color values, to obtain images with HDR effects. It works well in lit or backlit situations.
In addition, by applying the '200MP Adaptive Pixel' technology, the input image has three Color Filter Array (CFA) pattern structures of Tetra2 / Tetra / Bayer. They are integrated with AI demosaicing and denoising algorithms to maximize the detail from input images with different CFA structures.
A model of the sensor’s physical noise characteristics was made to create tens of thousands of high-resolution images for the AI neural network to train with. For more information on AI demosaicing and denoising, see Deep-Learning driven image quality-enhancing AI Te... - Samsung Members.
This gave us images with an extended dynamic range, maximizing detail while eliminating noise.
Galaxy S23 camera’s high resolution
By combining the above-mentioned 'Multi-frame image processing technology' and 'Deep Learning-based image quality-enhancing AI technology', we formed an optimal S/W ISP pipeline. We then applied it to the Galaxy S23 series to overcome the physical limitations of the sensor. We also introduced the '200MP Adaptive Pixel' technology and commercialized it to provide the best image quality on the Galaxy S23 Ultra.
Link with Camera Assistant
High-resolution shots can also be tied to Camera Assistant. As you can see in the image below, there are two options to choose from: ‘Prioritize Speed’ and ‘Prioritize Quality’.
'Prioritize Quality' applies the Multi-frame Image Processing with Neural Network technology mentioned above, to provide optimal image quality. However it takes time to get the final result because it requires software processing of many pixels. On the other hand, 'Prioritize Speed' applies conventional hardware image processing technology, and the image quality is slightly lower than 'Prioritize Quality', but the advantage is that you can check the results quickly.
It's important to note that when we talk about speed, we're not talking about "how fast you can take a series of shots". Instead it’s about "how fast you can get the results".
The process of capturing a single image can be roughly divided into two phases: 1) the capture phase, where you obtain the image to process, and 2) the processing phase, where you process the image. As you can see in the figure below, the ‘Prioritize speed’ mode handles capture and processing, one after the other. Conversely, the 'Prioritize Quality' mode captures images, one after the other, and separates the time taking process of ‘processing’ into a post-processing action. Therefore, the 'Prioritize speed' mode appears to be 'slower', but it is a 'faster' mode that allows you to see the results of your captures right away.
CamCyclopedia Index - Samsung Members
You can also check out CamCyclopedia at any time by going to Community -> Category (app) -> CamCyclopedia -> "CamCyclopedia Index".
4 Comments
