Algorithm to combine low-light and noisy video frames into an enhanced image.
Applications include thermal imaging, energy loss detection, night-time surveillance, night vision systems, and passive low-signal imaging systems.
Thermal images typically appear grainy and ill-defined because of their long wavelength and low signal strength.
This invention is an image-processing algorithm that recovers information from extremely noisy, low-light video and creates a high-resolution image. The program combines a series of low-resolution images intelligently by identifying their positions inside high-resolution reference images. The resolution of digital images is enhanced by moving the camera (dither-type) and estimating the point-spread function (PSF) from the motion. The PSF is conventionally assumed to be Gaussian and is applied to the image to reduce noise, but the Gaussian estimate is very uncertain. This is a much more robust approach and does not require calibration of the imaging system.
- Recover information from extremely noisy, low-light video
- Create super resolution image