A low-light image denoising enhancement method based on physical noise generation model

Abstract

The invention discloses a low-light image denoising and enhancement method based on a physical noise generation model, which belongs to the field of image processing. The present invention is applied to the low-light images captured by short-exposure in the low-light environment. According to the physical mechanism of the imaging sensor, the electronic imaging process of photons, current, voltage and digital signals is considered, and the physical noise generation model is constructed, using offset frames and flat fields. Frame, estimate the noise parameters of the physical noise generation model, correct the physical noise generation model, use the physical noise generation model to generate a large amount of low-light enhancement training data to train the neural network, and complete the high-quality noise reduction enhancement of the low-light image through the trained neural network. The invention can be used in imaging devices such as single-lens reflex cameras and mobile phones to reconstruct clear night scenes in the form of short exposures. It is suitable for occasions that require fast low-light imaging and expands the application range of low-light images. The invention can be used in the fields of video monitoring, high-speed photography and astrophotography.

Description

technical field [0001] The patent of the invention relates to a low-light image denoising and enhancement method based on a physical noise generation model, in particular to a method capable of reconstructing a high-quality clear and bright image from a low-light image captured by a short exposure in a low-light environment, which belongs to the field of image processing . Background technique [0002] Light is crucial to photography. Due to the limited number of photons and the unavoidable noise in the imaging process, night and low light place very strict requirements on photography. The traditional solution for low-light photography is to gather more light sources, including increasing the aperture, prolonging the exposure time, and turning on the flash source. However, each method has its own inherent flaws: large apertures result in small depth of field and are generally unusable in smartphone cameras; long exposures can introduce blur caused by scene changes or camer...

AI Technical Summary

Problems solved by technology

[0005] Aiming at the problem that the existing noise model ignores the physical characteristics of the imaging device and thus is not realistic enough, the technical problem to be solved by the low-light image denoising enhancement method based on the physical noise generation model disclosed by the present invention is: from the imaging sensor Starting from the mechanism of physical imaging, considering various noise sources in the imaging process, establishing a physical noise generation model, generating a large number of pairs of realistic data for training neural networks, and realizing noise reduction and enhancement of low-light images through the trained neural network, Extending the range of applications for low-light images

Images