Full text loading...
-
An Improve Convolutional Auto-Encode Denoising Method
- Publisher: European Association of Geoscientists & Engineers
- Source: Conference Proceedings, 82nd EAGE Annual Conference & Exhibition, Oct 2021, Volume 2021, p.1 - 5
Abstract
We evaluated an improve Convolutional Auto-Encode method for seismic data denoising. The method learn extremely complex functions to effectively attenuate noise by learning and extracting features from a large amount of training data set based on statistical techniques. However, the large quantity of training point pairs may increase the burden of memory and computation during the training. To solve the problem, we develop entropy sampling to select the effective training point pairs and reduce the training set based on the texture complexity. That is, complex texture regions represent the dominant characterization of the seismic data, and these regions are sampled with higher probability as training data set. Numerical illustrations on 2D seismic data show that the proposed method reduces the training data pairs as much as possible to improve the efficiency of training, while ensuring accurate denoising results.