portion of an image, need to remove pixels inside the ring and get pure brown color

I am processing a LandSAT image for geological purposes. After some processing there are so many sandy dots in the image. To get clear geological formation as a polygon I want to remove small sandy pixels. After 'majority' filter, by 3x3, 5x5, 9x9 many of them still remaining.

If you have any better idea for how to remove them the sandy, please share!


so i guess this is a classified image?

i can think of two more ideas :

1. i think for each class you can (start from biggest to lowest class) you can do morphological operations - open -> close (dilate -> erode).

2. treat them as polygons, find holes and remove them

you can do this for example in matlab

| improve this answer | |

You need to define the Minimum Mapping Unit (MMU). Also, treating 'clusters' of classified regions as polygons can help. as you are in Imagine I will provide software specific directions

  • apply focal majority filter (makes image polygon-like)

Raster>Thematic>Neighborhood>Majority (I like the 5x5)

  • 'Clump' the filtered image

Raster>Thematic>Clump (ignore zero in output stats)

  • sieve the results (This is what you where really looking for)

Raster>Thematic>Sieve -- Specify a minimum size of region to retain (i.e. 25 pixels) click OK!

Now, you will probably want to recode the image.

Happy Mapping!

| improve this answer | |

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.