That is a reasonable starting point. It won't work as well as you might think and that is where the creativity comes in.
Histogram will work pretty well in finding sunsets, snow scenes, forests and city scenes. That might be good enough for a student project. On Mon, Oct 4, 2010 at 4:50 AM, gagan chhabra <[email protected]>wrote: > Thanks a lot to you all. I have started going through the links you all > provided. > UCI seems like a crack to me now, but whatever happens will post here. > > Now, consider a set of images (P) or database and user inputs a query > image(Q) > > I got some approcach, its like- > > 1> the set P can be partitioned in { P1 , P2....Pn },virtually,on basis of > parameter like histogram color or histogram layout. > > 2>Obviously: P1 U P2 U P3...U Pn = P. Also it may be possible that : P1 > intersection P2 intersection...Pn != NULL set, it is because > some images may lie in more than one pattern or histogram oberved pattern. > > 3> the set P is partitioned on basis that images having same histogram > patterns can be listed in one set. > > 4>this way for one histogram patterns say *hp1 * there will be list of > images and there can be other sets as well. > > 5> Now the partitioned images observed pattern can be saved to compare with > Q. > > 6> Now the query image Q can be compared with patterns and the images of > matched pattern will be returned to user. > > > Is it ok... or should i give more thoughts to it..?? Any drawbacks or if i > missed out anything then please mention. > > > > > > On Mon, Oct 4, 2010 at 1:34 AM, Ted Dunning <[email protected]> wrote: > > > Try this: > > > > > http://www.public.asu.edu/~huanliu/sbp09/Presentations/paper%20presentations/SBP09_3-31(Baoxin%20Li%20-4).pdf > > > > On Sun, Oct 3, 2010 at 12:57 PM, Federico Castanedo < > [email protected] > > >wrote: > > > > > Hi Ted, > > > > > > Could you, please, post again the reference of the first paper. > > > > > > > > > > > > > > > -- > gagan >
