Skip to main content
22 events
when toggle format what by license comment
Apr 28, 2017 at 7:14 vote accept Muhammet Ali Asan
May 13, 2016 at 1:50 history bumped CommunityBot This question has answers that may be good or bad; the system has marked it active so that they can be reviewed.
S Sep 15, 2015 at 20:12 history suggested Laurent Duval CC BY-SA 3.0
Corrected text and added non-linear tag
Sep 15, 2015 at 18:34 answer added Laurent Duval timeline score: 2
Sep 15, 2015 at 17:53 review Suggested edits
S Sep 15, 2015 at 20:12
Aug 16, 2015 at 16:01 answer added Tolga Birdal timeline score: 1
Aug 16, 2015 at 15:32 comment added user17019 is their exist mean weighted median filter ?
Apr 1, 2015 at 11:15 answer added Batman timeline score: 0
Mar 31, 2015 at 7:46 history edited Muhammet Ali Asan CC BY-SA 3.0
added 16 characters in body
Mar 31, 2015 at 7:45 comment added Muhammet Ali Asan Weights are calculated over window not whole image.Weights can change if you slide the window. Weight function can decide weights.For example you may assign large weights to closest pixels to center.Then using the equation above you can find weighted median in window.
Mar 31, 2015 at 6:53 comment added Moti So, actually you are looking at an order rank (I assume that the weights are not changing as you move your window over the image) and instead of taking the middle one you are taking the one that fits the weight threshold. Are the weights distributed all over the image or only in a window? It seems that you do not deal with white noise.
Mar 30, 2015 at 20:57 comment added Niki Estner @Moti: I admit it's a bit confusing if you're thinking in terms of linear filters. It's not a linear filter, the weights aren't multiplied with the values. Intuitively, you sort all the pixel values, then you take the weight of the smallest value, plus the weight of the second smallest, ... plus the ith smallest value, until that sum is greater or equal to half of the total weight. then the ith value is the value you're looking for. If all the weights are 1, you get a median.
Mar 30, 2015 at 18:45 comment added Moti @nikie As far as I understand the process the weights are used to multiply the values of image pixels and then the median is found. Where the sum of weights comes from?
Mar 29, 2015 at 20:46 comment added Niki Estner @MuhammetAliAsan: Not really an answer, but this book contains a chapter on weighted median: [www.amazon.com/Handbook-Processing-Communications-Networking-Multimedia/dp/0121197905]. I think the common use case is to set the weight of the center pixel to e.g. N-1, all other weights to 1 - meaning "return the original pixel value, unless it's the highest or lowest value in the neighborhood. Then return the second highest/lowest value." This still removes salt&pepper noise, but leaves the image mostly unchanged otherwise.
Mar 29, 2015 at 20:43 comment added Niki Estner @Moti: There is no sum in median filter, but there's a count: $(n-1)/2$ values are lower then the result, the others are higher. Turn that count into a sum of weights, and you get the sum formula in Muhammet's question.
Mar 29, 2015 at 17:57 comment added Muhammet Ali Asan This is not a median , this is WEIGHTED MEDIAN. If all weights are 1 then it simply becomes traditional median filter.
Mar 29, 2015 at 15:34 comment added Moti I see the weights. Where are the image values? There is no sum in median filter (to the best of my knowledge) - this is usually an order filter that you pick a value that is the middle of the list. Weights may be used to remove certain bias, such as a slop background.
Mar 29, 2015 at 9:23 history edited Muhammet Ali Asan CC BY-SA 3.0
added 388 characters in body
Mar 29, 2015 at 9:13 comment added Muhammet Ali Asan @Moti that's what I am asking,spesific images. I am updating question.
Mar 29, 2015 at 5:33 comment added Moti You need to provide some context to your question - what is the problem you face, specific - image, videos, audio, sonar....?
Mar 29, 2015 at 3:58 comment added robert bristow-johnson well, i'm not into image processing (mentioning "pixel" hints at that), but i know what a median filter is. thing is, i dunno what a weighted median filter is. weighting samples in a mean makes some sense but i dunno what the sense is if you change the values of some samples, relative to others, before essentially sorting them. what is the rhyme or reason that sample $x[n_1]$ should be boosted over sample $x[n_2]$ when otherwise $x[n_1]<x[n_2]$? when sample order is swapped or changed, why are some samples selected and not others?
Mar 29, 2015 at 1:07 history asked Muhammet Ali Asan CC BY-SA 3.0