Avoid pixel overlaps while extracting separate tile images

Your place to discuss general handling of images and the CZI file format
Post Reply
Fredrik Olsson
Posts: 16
Joined: Mon May 25, 2020 9:53 am

Avoid pixel overlaps while extracting separate tile images

Post by Fredrik Olsson »

Hi,

This is a rather basic question but with great importance for us. There are some routines for extracting separate tile images from a large tiled image (for example there is an included job for that in Zen Core). My question is if these extracted tile images are similar to those individual images making up the tiled image in the beginning, which then may contain a stitching overlap, or if they are truly individual regions in the large tile image with no pixels in common? We want to make sure we do not analyze the same feature twice when we make our analysis on extracted individual tile images.


BR
Fredrik Olsson
CarlZeissMicroscopy3
Posts: 180
Joined: Wed May 20, 2020 10:10 am

Re: Avoid pixel overlaps while extracting separate tile images

Post by CarlZeissMicroscopy3 »

Hello Fredrik Olsson,

a tiled image is composed entirely of tiles. The acquisition itself or the image processing function ‘stitching’ register the tiles so that they overlap in the best way. It is possible to reset the stitching and get the original acquisition positions of the tiles. Finally it is possible to stitch them again to get a better overlap than before.

In other words, a tiled image is a collection of tiles unless the tiles are fused. By fusing a tiled image, a ‘normal’ image is obtained.
Extracting separate tile images from a large tiled image returns all the tiles individually. In any calculation etc. the overlaps must be considered.

When using Zen Analysis, Measurement or (most of the) Image Processing Functions the overlaps are treated internally so that the results are the same as if the image has been fused to a ‘normal image’.

I hope this helps.
Laydi Sary
Posts: 1
Joined: Thu Jun 22, 2023 11:46 am

Re: Avoid pixel overlaps while extracting separate tile images

Post by Laydi Sary »

the extracted tile images from Zen Core are similar to the individual images that make up the tiled image in the beginning, which may contain a stitching overlap. This means that there is a possibility of analyzing the same feature twice if you use the extracted tile images for analysis word games. One possible solution is to use a custom naming scheme for the export of individual tiles that includes the row and column information of each tile within the original tiled image. This way, you can avoid duplicating the analysis of overlapping regions. Alternatively, you can perform the analysis on the large tiled image directly using Zen Core or other software tools that support tiled image analysis Octordle
Farland Johnson
Posts: 2
Joined: Wed Aug 02, 2023 12:13 pm

Re: Avoid pixel overlaps while extracting separate tile images

Post by Farland Johnson »

Fredrik Olsson wrote: Tue Dec 28, 2021 6:21 pm Hi,

This is a rather basic question but with great importance for us. There are some routines for extracting separate tile images from a large tiled image (for example there is an included job for that in Zen Core). My question is if these extracted tile images are similar to those individual images making up the tiled image in the beginning, which then may contain a stitching overlap, or if they are truly individual regions in the large tile image with no pixels in common? We want to make sure we do not analyze the same feature twice when we make our analysis on extracted individual tile images.

Securitas ePay Login
BR
Fredrik Olsson
The answer to your question depends on the specific implementation and settings used for extracting separate tile images from a large tiled image.

In general, when tiling a large image, the tiles are often extracted in a way that there is some overlap between adjacent tiles. This overlap is commonly referred to as the "stitching overlap." The stitching overlap serves two primary purposes:

1. **Avoid Artifacts:** When processing or analyzing the tiled images, the stitching overlap helps avoid artifacts that may occur at the edges of tiles. These artifacts can be caused by various factors, such as filter effects, compression, or other image processing operations. The overlap ensures that the information from neighboring tiles is present, reducing potential discrepancies and errors at the tile edges.

2. **Seamless Transition:** The stitching overlap allows for a smooth transition when assembling the tiles back into the original large image. When reconstructing the large image, the overlapping regions are blended together, minimizing visible seams or lines where the tiles meet.

However, the size of the stitching overlap can vary depending on the specific application and settings used during the tiling process. In some cases, the overlap may be set to zero, resulting in no common pixels between adjacent tiles. This would mean that each tile represents an individual region with no shared pixels from its neighboring tiles.

To ensure that you don't analyze the same feature twice in your analysis, it's essential to understand the tiling process used and the size of the stitching overlap. If there is an overlap, you should consider analyzing only the non-overlapping regions of the tiles to avoid duplication. Alternatively, you may choose to analyze the entire tiles, including the overlapping regions, and account for any duplicated features in your analysis.

If you are using a specific software or library for tiling the images (such as Zen Core), it's best to consult the documentation or settings to determine the behavior of the tiling process and the size of the stitching overlap. This will help you accurately design your analysis workflow to avoid any duplication of features and ensure accurate results.
Post Reply