Prev: GUIDE Toolbar layout limitation
Next: 64 bit integer
From: Azy on 26 Jul 2010 05:41
HI,

I have a got a image of containing red cells and white blood cells.
http://drop.io./cellpic

Now in that image i want to count objects i.e circular objects in red excluding a big object in purple colour . So i tried converting it into a gray image. But when i converted into gray image even the big object was also converted .still i continued, I tried detecting edges using sobel. Then tried to close the boundaries at the edges and filled it with imfill function. by using bwarea() of one object and area of total image ,I tried to get the no of cells. but instead in process I forgot even im counting purple object. so thought i should change my plan , i need to try my hand on colour images. But coming to colour images its totally different .its not as easy as gray or B&W images.
SO now my aim is :

1) by retaining the color of image removing purple object from image.

2) identify red objects edges and boundaries

3) count how many red objects are there..

Hope by solving these lot of people may be benefitted who want to know about color images.. SO plz help me out guys....
From: ImageAnalyst on 28 Jul 2010 12:05
Try adapting my color detection demo:
http://www.mathworks.com/matlabcentral/fileexchange/authors/31862
From: Azy on 30 Jul 2010 06:54
ImageAnalyst <imageanalyst(a)mailinator.com> wrote in message <d8f04da3-8355-44c8-ae77-4b786d9d76b9(a)k19g2000yqc.googlegroups.com>...
> Try adapting my color detection demo:
> http://www.mathworks.com/matlabcentral/fileexchange/authors/31862



Well, ihave just chnage some threshold and aplied to my image.. but for some images it doesnt work.

http://drop.io/cellpic

if you go through above link to images ,for the last three it doesnt work. means apart from recognising red parts ,it is also recognising purple parts !!!!
From: ImageAnalyst on 30 Jul 2010 07:38
Perhaps you need to convert to HSV color space and use one of those
channels, or a combination of two or three channels to get what you
want. You need to figure out what differentiates the red from the
purple. If your 3D color gamut is not clustered along the RGB axes,
but goes at some weird angle, then you may need to use an alternate
color space, such as HSV, or even a customized one using principal
components analysis or linear discriminant analysis.
From: ImageAnalyst on 30 Jul 2010 07:48
Or maybe you could follow the MATLAB example
http://www.mathworks.com/products/demos/image/color_seg_k/ipexhistology.html
It discriminates red and purple objects.
 | 
Pages: 1
Prev: GUIDE Toolbar layout limitation
Next: 64 bit integer