Figure 1. Shows affected vs unaffected area with 60 metres
interpolation and 30 metres extrapolation.
This extrapolation/interpolation (JPG 58Kb)
was the default processing
for SLC-Off images processed before 31 December 2006.
The SLC-Off impacts are most pronounced along the edge of the scene and gradually diminish toward the centre of the scene as illustrated in Figure 1.
The total loss of image data has been estimated to be approximately 25% over any given scene. The maximum width of the data gaps along the edge of the image, or duplicated data in the case of Raw data and Raw image, would be equivalent to one full scanline, or approximately 390 to 450 metres. Geoscience Australia's processing system scales Landsat 7 SLC-Off products from 1 to 255, with a value of 0 DN used to black fill the scan gap.
The location of earth surface features not imaged will vary for any two scenes of the same area and the scan gap (black-fill pixels) between different bands is not identical. There is an offset for the coverage of the unaffected areas between different bands in the one scene. This offset is three kilometres between panchromatic bands and visible near infrared/shortwave infrared bands, and six kilometres between panchromatic band and thermal bands (JPG 144Kb). These gaps will not be visible on the browse image preview when searching and ordering an SLC-Off data product. The effect of the SLC-Off mode of acquisition on the final Landsat 7 data product will vary according to the level of correction applied.
| (a) Raw data (USGS Level 0R equivalent) and Raw image (USGS Level 1R equivalent) | |
Figure 2. Raw image product, showing duplication along the scene edge. |
Figure 3. Raw image product, showing nominal image data at the scene centre. |
| (b) Systematic products (Path or Map orientated image) and Precision products (Orthocorrected Image) | |
Figure 4. Orthocorrected product, illustrating the data gap along the scene edge. |
Figure 5. Orthocorrected product, showing nominal image data at the scene centre from interpolation/ extrapolation. |
Figure 6. False colour composite SLC-Off image. |
The scan gap shifts position slightly with spectral bands, resulting in valid data in some bands and no data in other bands. When the imagery was displayed as a colour image, individual pixels near the edge of the scan gap may appear as missing one or two bands. The red pixels in the example in Figure 6 show the presence of pixels with valid data in band 3 and no data in the other bands. The yellow pixels show the presence of pixels with no data in band 5 and valid data in the other bands. |
In January 2007, Geoscience Australia changed the default method of processing SLC-Off scenes in response to customer demand:
|
|
Figure 7. Interpolation/extrapolation (left) and |
Prior to 31 December 2006, our processing systems replaced the duplicated image data with null values (zero fill) and applied a small interpolation (60 metres) and extrapolation (30 metres) along the scanline boundary (jpg 58Kb). This process resulted in the middle of the scene, approximately 22 kilometres, containing very little duplication or data loss. This region is very similar in quality to previous Landsat 7 image data. The image in Figure 7 shows the comparison between SLC-Off products processed with and without interpolation/extrapolation near the centre of the scene. (Select for higher resolution 38Kb_image) Visual inspection suggests that noticeable impact of interpolation and extrapolation on SLC-Off products are observed on linear features. |
Figure 8. Panchromatic SLC-On (a) and SLC-Off (b, c and d) bands show |
The image in Figure 8 shows the disappearance of a road into a forest as the result of the interpolation applied on the image (panchromatic band, 60 metres, or 4 pixels interpolation). (Select for higher resolution 36Kb_image) |
Figure 9. Panchromatic SLC-On (A) and SLC-Off (B, C, D and E) bands show the effect of extrapolation on the direction of a linear feature with various resampling methods. |
The extrapolation process may distort the direction of a road as shown in Figure 9. The directions of the road from above and below the scan gap shift right and left respectively as a result of this process. (Select for higher resolution 31Kb_image) |
Figure 10. Duplicated pixels caused by extrapolation with NN resampling method. |
Products generated with NN resampling method shown in Figure 10 will have duplication of pixel values caused by extrapolation (30 metres) along the scanline boundary. (Select for higher resolution JPG 30KB) |