Hi,
I have conducted an unsupervised classification on the Landsat-5 image with 70 classes. Some
pixels are uniquely distributed and some are not. I want to clip out the pixels that are not appropriately distributed (commonly refers as mixed pixels) for fuzzy classification to ensure higher accuracy. So, I am trying to find ways to clip out the mix pixels from the unsupervisely classified image to a new file or layer if possible. I would be happy if there are some spatial models to do this.
References would be appreciated.
Thanks.
Masking is a great way to do this. It allows us to mask out areas depending on classes from the attribute table.
This option is available under the raster tab in Erdas-2014 and similar in other versions I suppose.
Click Raster tab > 📷 Subset & Chip 📷> Mask to perform this operation. Input the file you want to mask at the top left corner and input the file you want to mask according to; at top right corner. Then run setup recode and set the value as 0 for the classes you want to mask out from the image.
I hope it helps to whom are having trouble.
Thanks.
If some pixels are showing same color code then you will have to Re-classify the raster for that specific classification.
Thanks Ismail Mondal , it helps. I think it's the easiest way to do this. I would be thankful if you also take a look on my answer above.
- Badges
- Science topic
- Similar topics
- Computer Science
- Images