Table 1
Illustration of the detailed steps of the proposed method.
| Inputs: Model well log Xh, test low resolution log Yl, local patch size r, sampling interval δ, and neighborhood size K. |
| Step 1: Constructing model data pairs (Ml, Mh): |
| Step 1.1: Down-sampling Xh to obtain the corresponding low resolution well log Xl; |
| Step 1.2: Using (low-resolution) patch size r and (low-resolution) sampling interval being δ to continuously sample log patch pairs (Ml, Mh) from Xl and Xh; |
| Step 2: Initialing the expected super-resolution output Yh and the corresponding counting vector Ch to zeros. |
| Step 3: Conducting LLE-based super-resolution operation to Yl: |
| For k = 1: δ: length (Yl) |
Step 3.1: Obtaining the current local patch  ; |
| Step 3.2: Finding the corresponding K nearest neighbors from Ml; |
| Step 3.3: Calculating the LLE-based weight coefficient vector Wk using equation (7); |
Step 3.4: Estimating the corresponding super-resolution local patch  using Wk and Mh. |
Step 3.5: Accumulating  to Yh’s corresponding indexes  . |
Step 3.6:  ; |
| End |
| Step 4: Calculating the final Yh: |
| Yh = Yh/Ch (8) |
| Outputs: High resolution |