Pixel Gap in Source Rasters

Question

I am trying to build a mosaic dataset from tiled source rasters (.tiffs).  The problem is that there are two places with a 1 pixel gap across the entire set of rasters running horizontally.  It looks to be related to precision issues when they were generated.

Is there a way to sample the nearby pixels from the adjacent rasters and use the resampled value to fill in the gap?

Thanks,

Nathan

NathanHeick · Answer

To explain a little more the image sizes aren't in even ft.  The number of pixels are such that the height and width of the image in feet ends in .66 ft or .33 ft.

NathanHeick · Answer

The gaps are between tiles, not within tiles.  I see what you are saying if the gap is within the tile.

I think the problem is related to the cell size.  The images are supposed to be 4" resolution and the cell size says 0.33 ft.  The extents also seem to have precision down to 0.01 ft.  The leftmost column of images start at an even coordinate.  I think at some point the 1/300 (0.00333...) ft that is missing caused a pixel not to show up.  The coordinates in the extents will run like 0.66, 0.32, 0.98, 0.64, 0.30, 0.96, ... .

I am just wondering if there is a way to make the mosaic dataset sample from the pixels that are around the missing row and display the data in the final mosaic.

RobertBorchert · Answer

There is a possible chance your tiff was made from a compressed image format and the those pixels are the result of being uncompressed.

You may be able to fix it in something like photo shop.

Or possibly bring it into Picassa (copy  of image please) and try the I feel Lucky button.  Either way you are probably going to lose image quality without manually going through the broken pixels and manually adjusting them.

Once images are broken it is really hard to fix them.  Not impossible just not rewarding.

I am trying to build a mosaic dataset from tiled source rasters (.tiffs).  The problem is that there are two places with a 1 pixel gap across the entire set of rasters running horizontally.  It looks to be related to precision issues when they were generated.

Is there a way to sample the nearby pixels from the adjacent rasters and use the resampled value to fill in the gap?

Thanks,

Nathan