Displaced images when using tile configuration for stitching

Hi

I have been trying to stitch some bright-filed images using “Grid/Collection stitcing” in FIJI. In this plugin I have used “Positions from file” and the order is “Defined by TileConfiguration”. The tile configuration values are directly extracted from image metadata. Seems this method is working fine for a small group of images ( < 100Mb fused image) but not big grids of images. If I don’t check the “compute overlap” checkbox, the the image tiles won’t be in exactly correct place and if I use the “compute overlap” some of the tiles go missing and some would be displaced. Would you please give me some advice on this issue?

Thanks
Adib

Hi, tiles are dropped if all correlations with neighbouring tiles is lower than the specified threshold. Is it possible that this is the case? Can you maybe provide an example dataset where it fails?

Hi Stephan

Thanks for the reply. I have uploaded 207 raw images with their corresponding stitched image with tile configuration method: https://uwmadison.box.com/s/rrzngw2452auwei2rnxvr78i0u9lw1lb

As you can see in the stitched image the tiles are missing both in the back ground (which is corrected and has no information) and also on the tissue. So I don’t really what is causing this issue. Is that possible to tell FIJI not to move the tiles more than a specified distance?

Thanks
Adib

2 Likes

Hi Stephan

I just wanted to let you know that one of our group members, Guneet Singh Mehta, has started working this. We will keep you posted on this and will greatly appreciate any advice you might have for debugging.

Thanks
Adib

1 Like

Would it make sense to issue explicit warnings (e.g., to the Log window) whenever a tile is dropped? It seems like a lot of people are surprised by the missing tiles.