MT Erosion doesn't work on larger stars?

[Nocturnal]

Hi,

after viewing Harry's video I figured I'd give this 'make my stars smaller' thing a go. I created a star mask and started experimenting with different MT Erosion settings. Soon my stars became smaller but only the already small ones. Big stars stayed virtually untouched. How do I make the MT work on larger stars? I went all the way to the largest element size but saw no appreciable difference in its effect. I did redraw the circle in the new element so that wasn't the problem I think.

Thanks for any tips. I'd like to bring up this topic at MWAIC so I need to get it right

[Carlos Milovic]

Have you checked the opacity of the mask over those large stars?

[RBA]

Wouldn't different structuring elements affect stars of different size??? The structuring element acts as a mask to the morphological transformation, so too small of a mask I'd guess it could affect the outcome? Yes? No?

Also like Carlos pointed out, check your mask.

Unrelated to the question but I also wanted to comment and ask...

Of course each operator works best depending on what you want to do, but I've found myself using more and more often the Morphological Selection rather than the others. That allows me to control the amount of erosion and dilation via the Selection parameter - which for the purpose of reducing star size I tend to slide over to the left (more erosion, less dilation). Is that a "smarter" use of the MT tool or is just plain stupid?

[Harry page]

Hi

I can just see your mask settings and I see the scale is set at three , I would think this would miss the large star , I have used up to a scale of 6

Also I tend to find that amounts less than 0.5 work best with more than one iteration if necessary

Harry

[Silvercup]

Hi:

Setting Scale to 3 in StarMask only include small stars. Try setting scale to 5. You can always do 2 StarMask with scales 3 and 5 (for example) and then merge them with pixelmath star_mask+star_mask1

Best, Silvercup

[Carlos Milovic]

Hi Rogelio

You are doing it quite fine. The selection value is equal to a percentile value in a distribution curve. So, as values smaller than 0.5 represent lower pixel values than the median, they work fine as star size reducers. Minimum sometimes is too aggressive, or may incorporate/amplify noise.

BTW, I use to perform from 7 to 10 iterations, and amount values of 0.1 to 0.25. This yields smoother results.

[Nocturnal]

Hi,

thanks for all the comments. Taking a closer look at the mask it's clear that the larger stars have fainter 'blobs' in the star mask so that would indeed explain the small amount of erosion. I'll play with the star mask parameters some more. As I've reported in the past I've found that the star mask too easily includes areas far outside the star. I'm still having trouble making the mask represent the stars more accurately. It least it seems a star with a width of say 4 pixels should not result in a 15 pixel mask.