A denoising method based on cyclegan with attention mechanisms for improving the hidden distress features of pavement

Among the several types of added attention mechanisms, the closest to the original image is the one that only added the channel attention mechanism. Compared with adding the Channel Attention mechanism, the waveform is complete and precise, but there are some signals whose strength is not as good as adding the Channel Attention mechanism. The denoising results (Fig. Figure 12 Denoising results of different attention mechanisms (a) Original image (b) Denoising results without attention mechanism (c) Denoising results using channel attention (d) Denoising results using spatial attention (e) Denoising results using channel attention and spatial attention (f) Denoising results of PCA (g) Denoising results of SVD (h) Denoising results of wavelet threshold. For the denoised image, the label of the original image actually has a large error, so AP with proposed method is lower than original image.