Sparse Naïve Bayes Base on Entropy Correlation for GPR Image Denoising

In the process of ground penetrating radar data acquisition, the GPR signal often contains impulse noise and has a low signal-to-noise ratio. However, the performance of traditional de-noising methods is reduced when dealing with this problem. Therefore, a denoising method based on sparse Naive Bayes is proposed, which includes two stages of noise detection and noise removal. In the noise detection stage, a sparse naive Bayes model is constructed based on the Gray-level Co-occurrence Matrix (GLCM) to detect noise points before denoising, and the definition is added to the naive Bayes model The sparse descriptors make it more accurate for noise detection. In the noise removal stage, median filtering is used to process noise points, and Wiener filtering is used to enhance non-noise points. Compared with the traditional filtering method, the results show that this method can achieve good denoising effect even under high-density noise conditions, has a higher peak signal-to-noise ratio, and at the same time preserves image details.