Prev: Solving a transcendental equation
Next: SVM For OCR, is it possible to use input methods similar to neural net ?
From: Kenneth Galea on 2 Jan 2010 12:50
ImageAnalyst <imageanalyst(a)mailinator.com> wrote in message <069c23e5-e130-4908-be8a-76adb3784e79(a)g31g2000vbr.googlegroups.com>...
> On Jan 2, 10:21 am, "Kenneth Galea" <k.ga...(a)hotmail.com> wrote:
> > Yes ok, you're right there was no need to seperate variable names. However I don't see the need for using a binary image instead of RGB image (which still works)!! What's the difference?? In my program I initially used the binary image which was then labelled and filtered (by eliminating labelled objects <= 5 pixels) where an RGB image was used to identify different objects (with different colours)....why do I need to go back to my initial binary image??
> > Thanks Kenneth
> ------------------------------------------------------------------------------------
> Kenneth:
> OK, tell me how you're going to find boundaries on a pseudocolored,
> labeled image.
>
> If you do it (and it can be done if you unwisely decide to attempt
> it), it won't be in a single-line call, like my call to bwboundaries.
> It will take dozens or hundreds of lines of code. There are several
> ways to find boundaries on a labeled image, but the easiest would be
> to just convert the image to binary before you start. But then if
> you're going to do that then why not just call bwboundaries() on that
> binary image?
>
> You say "an RGB image was used to identify different objects (with
> different colours)." Just to clarify, the identification of different
> objects was done by the labeling function which created an integer
> image where each blob's pixels was given a single number. So you have
> islands (blobs) where everywhere in it is 1, and it's 2 for the next
> blob, and it's 3 for the next blob. No colors involved so far - just
> integers. You can convert this to a color image where each blobs is
> pseudocolored with a different color using the function call
> coloredLabels = label2rgb (labeledImage, 'hsv', 'k', 'shuffle'); %
> pseudo random color labels
> This will color the blobs for display and in that sense identify them,
> but does not identify them in the sense of being used "to identify
> different objects." Just wanted to clarify that, if not for you then
> for others.
>
> Regards,
> ImageAnalyst

Hi
Thanks for your clarification...you are right ....that's what I really meant.
I tried using B = bwboundaries(L,8, 'noholes') where L is a labelled matrix and also tried B = bwboundaries(binaryimage,8,'noholes') and when I debug the coordinates of the object contours in variable B are exactly the same for both solutions. I know it's strange but I checked and re-checked :/

Thanks Kenneth
From: ImageAnalyst on 2 Jan 2010 14:58
OK then maybe they've built bwboundaries to just look at zero and non-
zero pixel values. If it works for you, then great.
First  |  Prev  | 
Pages: 1 2 3
Prev: Solving a transcendental equation
Next: SVM For OCR, is it possible to use input methods similar to neural net ?