Data denoising method and device

Abstract

The invention discloses a data denoising method and device. The data denoising method and device are used for performing denoising on a large-scale data set so as to improve denoising processing rate. The data denoising method and device are suitable for application scenes which have higher requirements for processing speed such as stream processing. The method provided by the embodiment of the invention includes the following steps that: data requiring denoising are adopted as input and are substituted into a model carrying total variation (TV) norms; transformation is performed on the model the carrying total variation (TV) norms according to preset variable decomposition strategies; reconstruction transformation is performed on the transformed model based on the alternating direction method of multiplier (ADMM); and the model which is subjected to the reconstruction transformation is solved through iteration computation, so that de-noised data can be obtained.

Description

technical field [0001] The present invention relates to the field of communication technology, in particular to a total variation data denoising method and device based on an alternating direction method of multipliers (ADMM). Background technique [0002] Currently, data noise is widespread. For example, the data obtained from the Sensor Network (sensor network) often contains information that is not only unusable, but also affects, or even Noisy data used to mislead the data analysis process. For example, with digital image data, noise can cause image blur and affect image quality. [0003] A commonly used idea for data denoising is to introduce a regularization term based on the inherent structure of the data, and the strength of the data noise can be reflected by the size of the total variation (TV, Total Variation). In the prior art, Split-Bregman is a flexible algorithm for solving non-differentiable convex minimization problems, especially efficient for problems wi...

AI Technical Summary

Problems solved by technology

[0012] However, the computational complexity of the existing method is usually high, which requires better computing power to support the calculation process

As the data set continues to expand, the required computing power also increases, so the existing algorithms are not suitable for large-scale data sets

In addition, some methods that do not require high computing power usually take a long time to converge to the final result

However, the stream processing system contains the similar scenarios described in the aforementioned background technology, and the system needs to be able to perform fast processing so that the data can quickly enter the next step of processing. The existing related methods cannot provide good support, which will lead to Affect the implementation effect of the entire stream processing system

Images