SimpleCV - Color segmentation and region growing

Question

I would like some help regarding the color segmentation. I'm trying to identify some types of blood cells and some of them are distinguished by color. I was already excel in some images, the problem is that as the pictures do not have the same quality, are not equal, the colors change. I wanted to know if has as to do something like set a range of colors... Or if someone can explain to me a little better based segmentation regions growing using simplecv ... Thanks!

Answer 1

Given the range of colors it becomes more flexible to segment the image based upon that, but simultaneously it reduces the accuracy as more false positives are generated. You can relate it to the fact that when you increase the range you basically increase the allowance of error from your mean required color, which in turn can easily detect the regions which fall under the specified range but are not positive regions-of-interest.

The different approaches that can be implemented are:

1. Convert the image from RGB to HSV

2. Make use of morphology transformations.

   Why morphology transformations?

   • Because of the radiosity, adjacent objects under lightening have effect on their neighbors (like a green ball kept near red ball leaves a shade on the green color on red ball which can generate false positives easily, when objects are close to each other effect of this increases )