Prev: tutrial on time series deomposition with matlab
Next: fsolve Problem with 2 non-linear functions
From: sai on 3 Aug 2010 09:52
sir,

i will be having around 30 black and white pictures in tne database..
after that i will capture a b/w image and i need to compare it with the stored images.. and if there is any match i should be able to recognise that..

there is no need of checking the images by rotating them or something like that..
just overlap the images and find if they match for about 80 percent..

i want this to be done really fast as the system i am using needs to be really fast..

so what i have done is
by using sobel filter i was able to find out the edges of the images...

and next what should i do..

does finding out the edges and then comparing increase the speed of the system.???

finding the edges is not going to reduce the size of the image... so any suggestions plz

matlab will be running on the system and it should respond immediately..


From: us on 3 Aug 2010 09:54
On Aug 3, 3:52 pm, "sai " <u4ece08...(a)students.amrita.ac.in> wrote:
> sir,
>
> i will be having around 30 black and white pictures in tne database..
> after that i will capture a b/w image and i need to compare it with the stored images.. and if there is any match i should be able to recognise that...
>
> there is no need of checking the images by rotating them or something like that..
> just overlap the images and find if they match for about 80 percent..
>
> i want this to be done really fast as the system i am using needs to be really fast..
>
> so what i have done is
> by using sobel filter i was able to find out the edges of the images...
>
> and next what should i do..
>
> does finding out the edges and then comparing increase the speed of the system.???
>
> finding the edges is not going to reduce the size of the image... so any suggestions plz
>
> matlab will be running on the system and it should respond immediately..

show CSSM an example of your images...

us
From: ImageAnalyst on 5 Aug 2010 09:38
They all have different areas. How about just summing your binary
images to get the area and see which reference image comes closest to
that?
From: sai on 9 Aug 2010 00:37
ImageAnalyst <imageanalyst(a)mailinator.com> wrote in message <bb438ac6-3e06-4b0c-93d4-a04d1748728a(a)l20g2000yqm.googlegroups.com>...
> They all have different areas. How about just summing your binary
> images to get the area and see which reference image comes closest to
> that?

i have tried it out by summing the binary values of images..and checking which reference image value is closest .this didnot work for most of the cases...
please suggest any other method
From: ImageAnalyst on 9 Aug 2010 12:32
Try looking at other things, like the shape, etc. Compare the
perimeter^2/area ratio of the objects, or the aspect ratio of the
objects. You need to find something that will tell them apart. If
all else fails, just measure everything (using the'all' parameter in
regionprops) and maybe you can get some statistics help to determine
what are the unique things about each reference image.
 | 
Pages: 1
Prev: tutrial on time series deomposition with matlab
Next: fsolve Problem with 2 non-linear functions