Method and system for quickly realizing image noise reduction based on MPA&BIB3D

Abstract

The invention discloses a method and a system for quickly realizing image noise reduction based on MPA&BIB3D. In the first stage, different noise reduction methods (such as discrete transformation, wavelet transformation and the like), different partitioning scales and different parameters are adopted to carry out hard threshold noise reduction on an image needing noise reduction, a plurality of images with different noise reduction levels are obtained, then weighted average is carried out on all the images, and a temporary noise reduction image is obtained; in the second stage, the original image block is read block by block, then a plurality of big blocks (bb) containing the current block are generated, a denied bb (dbb) corresponding to the bb in the previous stage is found, the power spectrum of the dbb is solved, the Wiener coefficient is solved, frequency domain transformation is carried out on the bb, then the bb is multiplied by the Wiener coefficient to obtain a new frequency domain value, inverse transformation is carried out to obtain bb after noise reduction, and a noise reduction result corresponding to the current block is taken out; and a weighted average value of all the noise reduction results is solved as a noise reduction result of the current block, and a final noise reduction image after circulation is obtained.

Description

technical field [0001] The invention relates to the field of image processing, in particular to image noise reduction. Background technique [0002] The quality of the image preprocessing algorithm is directly related to the effect of subsequent image processing, such as image segmentation, target recognition, edge extraction, etc. In order to obtain high-quality digital images, it is often necessary to denoise the image and keep it as possible as possible. While maintaining the integrity of the original information (that is, the main features), it can also remove useless information in the signal. [0003] Image noise is divided into random noise and sensor noise. Random noise is easier to remove, while sensor noise is more difficult to remove. For example, thermal noise will be formed on the long-term exposure image of the imaging machine under the condition of poor light, resulting in low image quality. [0004] The currently commonly used image denoising algorithms can...

AI Technical Summary

Problems solved by technology

At present, the most excellent BM3D algorithm has the characteristics of denoising in the air domain and transform domain. Although this algorithm is excellent, it takes a long time and cannot meet the requirements of real-time applications.

Images