Table 5 Ablation study on the Test_L400 and Test_U90 datasets. Here, model-A represents only inputting \(y_0\) as the input condition, model-B represents inputting both \(y_0\) and \(y_0\)-\(x_t\) as the input conditions, model-C represents inputting both \(y_0\) and content compensation module as the input conditions, and model-D represents a full model.
From: CPDM: Content-preserving diffusion model for underwater image enhancement
Models | Test_L400 | Test_U90 | ||||
|---|---|---|---|---|---|---|
PSNR \(\uparrow\) | SSIM \(\uparrow\) | MSE \(\downarrow\) | PSNR \(\uparrow\) | SSIM \(\uparrow\) | MSE \(\downarrow\) | |
model-A | 24.01 | 0.87 | 0.0074 | 19.95 | 0.80 | 0.0164 |
model-B | 24.72 | 0.88 | 0.0063 | 20.31 | 0.80 | 0.0153 |
model-C | 24.26 | 0.87 | 0.0073 | 21.03 | 0.82 | 0.0129 |
model-D | 25.11 | 0.88 | 0.0056 | 21.07 | 0.84 | 0.0114 |
