X
PrevPrev Go to previous topic
NextNext Go to next topic
Last Post 04 Nov 2016 09:07 PM by  anon
L7 SLC-Off Mosaic + Gap Fill?
 1 Replies
Sort:
You are not authorized to post a reply.
Author Messages

anon



New Member


Posts:
New Member


--
04 Nov 2016 09:07 PM
    Hey guys. Another new guy with his second question in as many days. I'm using SLC-off data for a project. I'm trying to fill in the SLC gaps in two images of the same scene, 16 days apart (reducing temporal variation, crop rotation, etc). Both of the images, when cropped to my study area, are cloud free. Per the Gap Phase calculator tool, I should have 94% in my study area. I first used the ENVI Mosaic tool, as suggest on the USGS site. It worked fairly well, but I got what appear to be resampling errors where the SLC-off gaps used to be. When viewed in most band combinations, they'll drop what look to be some impervious pixels in the middle of a farm field, for example. These exist exclusively next to the SLC-off gaps. I'm using 0s for both of the Data Ignore Values. No other changes. I then used the Gap Fill tool. Works pretty well, but it's creating some obvious "streaking" in areas with spectral heterogeneity that aren't there in the two original SLC-off files. It works fine when going through a farm field or a forest, upon visual inspection, it drops off when I observe a boundary like a lake or the transition from rural to urban. It seems to me like the (Seamless) Mosaic is the better option. It's taking real data and splicing it together to fill in the gaps. Anyone have any suggestions on how to approach the problems expressed above?

    MariM



    Veteran Member


    Posts:2396
    Veteran Member


    --
    07 Nov 2016 12:09 PM
    I do think you will get the best results using mosaicking. Are you using ENVI 5.3.1? There were some updates to the seamless mosaic from 5.2 and 5.3. I noticed the help mentions some issues with scan lines in ETM data: Some Landsat-7 scenes may have scan line artifacts around the edge of the imagery. If you do not trim the bad pixels prior to using the data in Seamless Mosaic, you can get unexpected results when applying feathering and adding seamlines. Consider creating and applying a mask to exclude these pixels. It may be a similar issue with the SLC-off areas. Are you setting a feathering distance for the mosaic borders? The default is 0 but you could try something like 20 or 50 pixels. It should improve those edge areas but may change some pixel values somewhat in those areas as well.
    You are not authorized to post a reply.