Dark light image noise reduction method and device

Abstract

The embodiment of the invention provides a dark light image noise reduction method and device. The method comprises the following steps: acquiring a RAW domain dark light image; performing preset image enhancement transformation on the RAW domain dark light image, and inputting the transformed image into a pre-trained noise reduction network model to obtain an output image; wherein the noise reduction network model is trained according to sample images, and the sample images comprise a simulation dead pixel dark light image and a noise-free image; and performing inverse transformation of preset image enhancement transformation on the output image to obtain a noise-reduced image. The influence of image dead pixels on the dark light image noise reduction process can be remarkably reduced, and the noise reduction quality of the dark light image is improved.

Description

technical field [0001] The present invention relates to the technical field of image processing, in particular to a method and device for image noise reduction in dark light. Background technique [0002] With the rapid development of computer vision technology and optical imaging technology, the method of using video image acquisition equipment to acquire images or videos and process them has been widely used in various aspects such as security, coastal defense, and intelligent transportation. [0003] However, when the imaging device is under dark light conditions, due to the severe lack of light, the collected images will contain a lot of noise, making the images unclear and poor in quality, which in turn will reduce the accuracy of post-processing results. Therefore, such images need to be denoised before being analyzed and processed, and during the denoising process, the bad pixels in the image will seriously affect the image denoising quality. [0004] The imaging ele...

AI Technical Summary

Problems solved by technology

[0006]Although many Image Signal Processing (ISP) algorithms include the link of dead point correction, due to the diversity and complexity of dead points, Based on traditional image processing methods, it generally cannot be completely eliminated, and the intensity of removing dead pixels is relatively high, which will cause the image to produce a smoothing effect similar to the median filter, and damage the edge details of the image.

Images