Timeline for Median Filter one after another
Current License: CC BY-SA 3.0
28 events
| when toggle format | what | by | license | comment | |
|---|---|---|---|---|---|
| Feb 18, 2013 at 13:20 | history | tweeted | twitter.com/#!/StackSignals/status/303494141466255362 | ||
| Feb 16, 2013 at 20:36 | history | migrated | from stackoverflow.com (revisions) | ||
| Feb 16, 2013 at 9:50 | comment | added | thang | guys, there's a couple things to point out here. 1. this is a programming question and the moderator is an idiot. 2. there is a point in a single 5x5 operation even if it were just simply a cascade of two 3x3. this is because it is a single pass through the image and offers better cache localization than two separate passes. | |
| Feb 3, 2013 at 4:51 | vote | accept | Gilad | ||
| Feb 2, 2013 at 22:34 | answer | added | Gilad | timeline score: 2 | |
| Feb 2, 2013 at 15:45 | answer | added | sapwood | timeline score: 1 | |
| Feb 2, 2013 at 15:30 | answer | added | sega_sai | timeline score: 2 | |
| Jan 31, 2013 at 22:42 | comment | added | mmgp | @Androidy this isn't a rule for non-linear filters. I would try to explain how it works for certain non-flat morphological structuring elements, but from previous discussions about flat and non-flat morphology in SO I'm unsure if there is any point in doing so. | |
| Jan 31, 2013 at 22:35 | comment | added | Gilad | @mmgp i think there is no Mask that represents the median filter since it's non-linear as well you can do A*A(IM)=A^2(IM). | |
| Jan 31, 2013 at 20:03 | comment | added | mmgp | @OliCharlesworth I noticed I made a mistake in the example, I'm sorry for that, but at least we are in agreement that this function is doing nothing else than the two passes. Thus, nothing has been done to answer the question (I currently don't think it is answerable in a positive way). | |
| Jan 31, 2013 at 19:53 | comment | added | mmgp | @OliCharlesworth why are you fighting against the example ? First calculate a 3x3 median filter on it, pad with ones as needed. Now calculate a 3x3 median filter on the result obtained previously. The central point is 1 now. | |
| Jan 31, 2013 at 19:32 | comment | added | Oliver Charlesworth | @mmgp: For a 5x5 example, my "function" is literally identical to running two passes! I don't understand what you'd be doing to obtain a value of 1 for the central pixel. | |
| Jan 31, 2013 at 19:30 | comment | added | mmgp | @OliCharlesworth it certainly isn't, check the simple example. | |
| Jan 31, 2013 at 19:27 | comment | added | Oliver Charlesworth | @mmgp: Huh? By definition, what I'm describing is identical to just performing two monolithic passes! | |
| Jan 31, 2013 at 19:26 | comment | added | mmgp | @OliCharlesworth i.imgur.com/Ns0bbKs.png, the central point there is assigned 0, but with two 3x3 passes it would have the value 1 (the notations I've used for showing the medians of the windows might be confusing, don't let distract you and check it yourself if needed). | |
| Jan 31, 2013 at 19:14 | comment | added | Oliver Charlesworth | @mmgp: Something like that, yes. (I see now that I should have started my description with "given a 5x5 region...") ;) | |
| Jan 31, 2013 at 19:12 | comment | added | mmgp |
@OliCharlesworth while someone could call that as a 5x5 function, it is not one that produces the correct result for two 3x3 median filter passes. I could prepare a counter-example if that is needed at all. Or maybe I misunderstood your function. Did you mean the following: given a 2D discrete point (x, y), compute the median of the medians of the nine 3x3 windows around it, then assign the resulting median to (x, y). Was that it ?
|
|
| Jan 31, 2013 at 19:04 | comment | added | Oliver Charlesworth | @mmgp: Expressed like that, there isn't one. I was merely pointing out that a 5x5 function very definitely exists. | |
| Jan 31, 2013 at 18:54 | comment | added | mmgp | @OliCharlesworth huh ? What is the point of doing that then instead of doing two 3x3 passes ? That is not combining the two passes into one. | |
| Jan 31, 2013 at 18:52 | comment | added | Oliver Charlesworth | @mmgp: The function is: compute the 9 medians of all possible 3x3 regions, and then compute the median of those results. | |
| Jan 31, 2013 at 17:39 | comment | added | mmgp | @OliCharlesworth I would like to see it. | |
| Jan 31, 2013 at 17:36 | comment | added | Oliver Charlesworth | @mmgp: Sure. It won't be a pleasant operation, but fundamentally it's just a non-linear function of a 5x5 neighbourhood. | |
| Jan 31, 2013 at 17:25 | comment | added | mmgp | @OliCharlesworth I'm interested in that too, are you sure it can be done for the median filter ? In that case, can the root signal be found in a single step ? | |
| Jan 31, 2013 at 8:06 | comment | added | Gilad | @OliCharlesworth can you tell me how this can be done? | |
| Jan 31, 2013 at 8:00 | comment | added | Gilad | @sgar91 it's not 3 operations. it's 2 . first you do the first Median and after that on the result . you do the next one. please explain why does it matter that it's not linear? | |
| Jan 31, 2013 at 7:59 | comment | added | Oliver Charlesworth | @sgar91 well you can, but it wouldn't be as simple as the median of a5x5 window. | |
| Jan 31, 2013 at 7:58 | comment | added | sgarizvi | Median filter is non linear. You cannot combine multiple Median filter operations into a single operation. | |
| Jan 31, 2013 at 7:57 | history | asked | Gilad | CC BY-SA 3.0 |
