A method and device for image background reconstruction

Abstract

The present invention provides an image background reconstruction method and device. The method includes: obtaining two adjacent frames of images, respectively calculating the absolute value of the difference between the pixels in the left and right regions of the adjacent two frames of images, and counting the left and right regions. The total number of pixels whose absolute value is greater than the preset threshold is respectively used as the first total number of pixels and the second total number of pixels; according to the first total number of pixels and the second total number of pixels, respectively calculate the left area impulse and the right area impulse; the current image The product of the learning rate of the previous frame of the image and the impulse of the left region is used as the learning rate of the left region of the current image; the product of the learning rate of the image of the previous frame of the current image and the impulse of the right region is used as the learning rate of the right region of the current image; The preset reconstruction calculation formula, the learning rate of the left region and the learning rate of the right region of the current image reconstruct the pixel value of each pixel in the current image to obtain the background of the current image. By applying the embodiments of the present invention, the reliability of image background reconstruction is improved.

Description

technical field [0001] The present invention relates to the technical field of image processing, in particular to an image background reconstruction method and device. Background technique [0002] With the development of image processing technology, more and more fields need to reconstruct the background in the image. At present, there are two common background reconstruction methods, one is the model method, and the other is the grayscale normalization method. class law. The model method is to establish a statistical model (such as a single Gaussian model, a mixed Gaussian model, etc.) for the gray level of each pixel. The reliability is not high; although the grayscale classification method has a small amount of calculation, it does not consider the impact of scene changes. For example, due to the walking of people in the scene, the reconstructed background is prone to empty positions (also called "ghosting") ” region), resulting in suboptimal refactoring results. Ther...

AI Technical Summary

Problems solved by technology

The model method is to establish a statistical model (such as a single Gaussian model, a mixed Gaussian model, etc.) for the gray level of each pixel. The reliability is not high; although the grayscale classification method has a small amount of calculation, it does not consider the impact of scene changes. For example, due to the walking of people in the scene, the reconstructed background is prone to empty positions (also called "ghosting") ” region), resulting in suboptimal refactoring results

Therefore, the reliability of the background reconstructed by existing background reconstruction methods is not high

Images