Dear all, I have a question about adaptive filtering for image enhancement. I want to design an algorithm for image enchancement. It should work on most of images. So I need to make it adaptive. I guess what I need is to extract some features from different images. I did extract the largest eigenvalue of the covariance matrix of the input images. I use this eigenvalue as "x". Then I try to find the best filter for this image, and use that filter parameter as "y". Then I work on a bunch of images, and get a bunch of these "x"s and "y"s. And do curve fitting to find the relation between "x" and "y" and hopefully this curve can predict "y" for future testing unknown "x". I want to ask is this a common way of training an adative filter for image processing? What kind of techniques yuo guys are using? Could anybody point me to some resources? I have searched a bunch of image processing books, but they are not very useful for this topic... Thanks a lot, -Walala
what are the training algorithm you guys are using for adpative filtering for image processing?
Started by ●October 24, 2003
Reply by ●October 24, 20032003-10-24
walala wrote:> Dear all, > > I have a question about adaptive filtering for image enhancement. > > I want to design an algorithm for image enchancement. It should work on most > of images. So I need to make it adaptive. > > I guess what I need is to extract some features from different images. I did > extract the largest eigenvalue of the covariance matrix of the input images. > I use this eigenvalue as "x". > > Then I try to find the best filter for this image, and use that filter > parameter as "y". > > Then I work on a bunch of images, and get a bunch of these "x"s and "y"s. > And do curve fitting to find the relation between "x" and "y" and hopefully > this curve can predict "y" for future testing unknown "x". > > I want to ask is this a common way of training an adative filter for image > processing? What kind of techniques yuo guys are using? Could anybody point > me to some resources? I have searched a bunch of image processing books, but > they are not very useful for this topic... > > Thanks a lot, > > -WalalaCome on: you should know better by now. Give your friends a break. Tell them what you want the filter(s) to do. Add detail in the shadows? Add detail in the highlights? Both? Alter gamma? Sharpen edges? Flatten illumination? Should it work on all or only on portions of the image? Black and white or color? A busy CEO once had a doctor come to his office so he wouldn't have to leave his desk to get his annual physical. When the doctor asked a question that interrupted the CEO's reading of the Wall Street Journal, the CEO said, "You're the doctor. You're supposed to tell _me_ how I am." The doctor packed his bag and said, "I'm going to refer you to a colleague who knows how to conduct a physical without asking questions. He's a veterinarian." I'm not a veterinarian. Jerry -- Engineering is the art of making what you want from things you can get. �����������������������������������������������������������������������
Reply by ●October 24, 20032003-10-24
"walala" <mizhael@yahoo.com> wrote in message news:bna5kv$890$1@mozo.cc.purdue.edu...> Dear all, > > I have a question about adaptive filtering for image enhancement. > > I want to design an algorithm for image enchancement. It should work onmost> of images. So I need to make it adaptive. > > I guess what I need is to extract some features from different images. Idid> extract the largest eigenvalue of the covariance matrix of the inputimages.> I use this eigenvalue as "x". > > Then I try to find the best filter for this image, and use that filter > parameter as "y". > > Then I work on a bunch of images, and get a bunch of these "x"s and "y"s. > And do curve fitting to find the relation between "x" and "y" andhopefully> this curve can predict "y" for future testing unknown "x". > > I want to ask is this a common way of training an adative filter for image > processing? What kind of techniques yuo guys are using? Could anybodypoint> me to some resources? I have searched a bunch of image processing books,but> they are not very useful for this topic... >Walala, You might try sci.image.processing for this question. Oh! I see you've already done that.... You didn't say what the objective criterion for the adaptive filter was going to be. Hard to advise in that case. What are you trying to do? i.e. what determines "the best filter"? Why is the largest eigenvalue of the covariance matrix of an image relevant to your choice of "the best filter"? Unless you have a physical justification for choosing this parameter then it may as well be any other measure. How about the sum of all elements of the image? etc. etc. Jerry Avins forwarded a message from you that had "attached files". Where are they? I'm not aware of much in the way of adaptive filtering for images. I guess it depends on what you're trying to do with the filter (which you haven't said). For example: - I guess you might have an adaptive filter that would warp an image with respect to a reference image in order to a distorting function in order to remove the distortion. I don't know how to do this but I do know that adaptive mirrors have been built in order to deal with atmospheric distortion. - You might use adaptive filtering to remove periodic noise in much the same way that it's done in 1-D signal processing. - You might use adpative filtering to enhance periodic structure in an image while reducing noiselike parts in the same way that it's done in 1-D signal processing. Thus, knowing what you're attempting to accomplish and your goodness criteria is quite important. At this point, my feeling is that all the discussion about extrapolation isn't so much to the point here - notwithstanding that the admonitions from others in this regard are good ones in general. I'm thinking you have a more fundamental problem. Fred
Reply by ●October 24, 20032003-10-24
"Jerry Avins" <jya@ieee.org> wrote in message news:bna9iu$mbv$1@bob.news.rcn.net...> walala wrote: > > > Dear all, > > > > I have a question about adaptive filtering for image enhancement. > > > > I want to design an algorithm for image enchancement. It should work onmost> > of images. So I need to make it adaptive. > > > > I guess what I need is to extract some features from different images. Idid> > extract the largest eigenvalue of the covariance matrix of the inputimages.> > I use this eigenvalue as "x". > > > > Then I try to find the best filter for this image, and use that filter > > parameter as "y". > > > > Then I work on a bunch of images, and get a bunch of these "x"s and"y"s.> > And do curve fitting to find the relation between "x" and "y" andhopefully> > this curve can predict "y" for future testing unknown "x". > > > > I want to ask is this a common way of training an adative filter forimage> > processing? What kind of techniques yuo guys are using? Could anybodypoint> > me to some resources? I have searched a bunch of image processing books,but> > they are not very useful for this topic... > > > > Thanks a lot, > > > > -Walala > > > Come on: you should know better by now. Give your friends a break. Tell > them what you want the filter(s) to do. Add detail in the shadows? Add > detail in the highlights? Both? Alter gamma? Sharpen edges? Flatten > illumination? > > Should it work on all or only on portions of the image? Black and white > or color? >If that's is so simple as "sharpening edge", "adjusting histogram", etc. , I may just need to grab a book and search for algorithm. But that's not. This is to get rid of small blocks in in image. Suppose you originally have a smooth arc in the image, but now it becomes ladders due to small blocks. I found the best filter for different images is different. So need to find an adaptive algorithm.
Reply by ●October 24, 20032003-10-24
walala wrote:> "Jerry Avins" <jya@ieee.org> wrote in message > news:bna9iu$mbv$1@bob.news.rcn.net... > >>walala wrote: >> >> >>>Dear all, >>> >>>I have a question about adaptive filtering for image enhancement. >>> >>>I want to design an algorithm for image enchancement. It should work on > > most > >>>of images. So I need to make it adaptive. >>> >>>I guess what I need is to extract some features from different images. I > > did > >>>extract the largest eigenvalue of the covariance matrix of the input > > images. > >>>I use this eigenvalue as "x". >>> >>>Then I try to find the best filter for this image, and use that filter >>>parameter as "y". >>> >>>Then I work on a bunch of images, and get a bunch of these "x"s and > > "y"s. > >>>And do curve fitting to find the relation between "x" and "y" and > > hopefully > >>>this curve can predict "y" for future testing unknown "x". >>> >>>I want to ask is this a common way of training an adative filter for > > image > >>>processing? What kind of techniques yuo guys are using? Could anybody > > point > >>>me to some resources? I have searched a bunch of image processing books, > > but > >>>they are not very useful for this topic... >>> >>>Thanks a lot, >>> >>>-Walala >> >> >>Come on: you should know better by now. Give your friends a break. Tell >>them what you want the filter(s) to do. Add detail in the shadows? Add >>detail in the highlights? Both? Alter gamma? Sharpen edges? Flatten >>illumination? >> >>Should it work on all or only on portions of the image? Black and white >>or color? >> > > > If that's is so simple as "sharpening edge", "adjusting histogram", etc. , I > may just need to grab a book and search for algorithm. But that's not. This > is to get rid of small blocks in in image. Suppose you originally have a > smooth arc in the image, but now it becomes ladders due to small blocks. I > found the best filter for different images is different. So need to find an > adaptive algorithm.How will your adaptive filter treat a partly finished brick wall? Will it reshape the bricks? Jerry -- Engineering is the art of making what you want from things you can get. �����������������������������������������������������������������������
