I am getting strange image artifacts when I co-register a large stack of Radarsat-2 images. I am using the co-registration module (create stack --> cross-correlate --> warp). I’ve played with some parameters including using the minimum extent of the images present in the stack and the warping functions (cubic, linear etc.) The images are all from the mode/viewing geometry so the co-registration should be fairly straight forward. Does anyone have any suggestions or an explanation for why this may be happening?
Yes, the uneven border. I have never seen that before. The reason why I became concerned with this is because this particular dataset generates an error in StaMPS at the last step (mean_v cannot be found). I wondered if it had something to do with the data because it works well on other datasets.
Yes, I thought that the image inputs for StaMPS shouldn’t be terrain corrected as well! However, the output points were shifted in space, and if the inputs are terrain corrected then the points have correct geolocation (i.e. the points overlap with man-made structures, roads etc.). Also, please see Katherine’s comment in this thread: Workflow between SNAP and StaMPS.
this is true, you can technically terrain correct the complex pixel information and work with it in StaMPS. If this increases the quality of your geolocation, you can proceed as Katherine listed it.
But be assured that the edges are not related to the error in StaMPS, this is common for all terrain corrected SAR images.