Super-Resolution

Based on multi-frame fusion along with the precise characterization of sensor and lens properties, Almalence Super-Resolution captures image details beyond the sensor pixel count and density and achieves higher SNR beyond the limit imposed by pixel size.

Unlike upscaling and sharpening techniques, creating an illusion of a more detailed picture, Almalence Super-Resolution actually reconstructs fine details (scientifically speaking – achieves MTF beyond the Nyquist limit of the sensor, or, simpler, captures the details smaller than a pixel), providing images and video which could be only captured with a higher resolution camera with bigger pixel size.

Left: the resolution limit of a camera;
Right: Same camera with Almalence Super-Resolution

Almalence Super-Resolution is used in:

  • Smartphones, to enable high-quality lossless zoom (our customers achieve the top DXOMARK Tele Zoom scores);
  • Medical cameras, to allow doctors to see more with the smallest endoscopes;
  • Machine vision cameras, to achieve robust objects recognition in adverse conditions;
  • VR/AR head-mounted displays, to enhance the picture quality of see-through cameras;
  • Laptop cameras, to achieve high picture quality whilst keeping thin bezel;
  • IoT, drones, and other industries – wherever the picture quality is limited by camera size, pixel count, or lighting conditions.

How it works

Almalence Super-Resolution simultaneously overcomes the limitations posed by sensor resolution capability, optical blur, and pixel noise using the following principles:

  • Registration of consecutive exposures with sub-pixel precision captures the details of a size smaller than the one imaged by a sensor pixel;
  • Neural network-based de-blur reconstructs the details beyond the optical blur and pixel noise.

Key features making Almalence Super-Resolution stand out and superior over other methods:

  • Non-iterative algorithm for short and predictable processing time.
  • Non-hallucinating AI reconstructs the image details, not guessing and replacing the actual picture with images from a dataset.
  • Algorithm architecture fit for vector processing enables real-time operation at video frame rates.
  • Precise calibration to achieve the topmost image quality at any given optical system.
  • Based on universal principles it is applicable to all kinds of cameras, from the tiniest endoscopy imagers to smartphone cameras to highest-end DSLRs.