Random Noise Attenuation Using Non-local Bayes Algorithm

Seismic random noise suppression using Non-local Bayes (NL-Bayes) algorithm is an improved algorithm of Non-local means (NL-means). NL-means algorithm uses a weighted mean of the most similar data patches in a neighborhood to replace each data patch. However NL-Bayes algorithm by evaluating for each group of similar patches a Gaussian vector model replaces a weighted mean of the most similar data patches in a neighborhood to improve NL-means algorithm. NL-Bayes algorithm is implemented in two identical iterations, and in the second iteration it uses the denoised seismic data of the first iteration to estimate better the means and covariance of the patch Gaussian models improving the similarity of data patches for achieving random noise attenuation. We explore its applicability to process industrial data sets via tests with synthetic and field data.