How does it work OpenCV Mser for color images?

asked 2016-03-25 07:58:00 -0500

Davit gravatar image

updated 2016-03-25 08:01:59 -0500

I am using OpenCV Mser class.
For grayscale images I don't have any problems, reults are good and parameters defined in documentation influences to algorithm logically :

  • delta - it compares (sizei−sizei−delta)/sizei−delta
  • min_area - prune the area which smaller than minArea
  • max_area - prune the area which bigger than maxArea
  • max_variation - prune the area have simliar size to its children

But I have problems with Mser for color images. I have looked to the source code and didn't find separate implementation for colored images(here is link of source code: https://github.com/Itseez/opencv/blob... ). The main important thing to understand if it finds regions by using threshold levels of an image, what is the equation for the "delta" parameter, it should be different from equation defined for gray images. If it uses another algorithm, the exact meanings of the parameters mentioned below :

  • maxEvolution; //! for color image, the evolution steps
  • areaThreshold; //! the area threshold to cause re-initialize
  • minMargin; //! ignore too small margin

because result of function for color images isn't descriptive, there are a lot of regions and it's hard to udnerstand how algorithm works.

edit retag flag offensive close merge delete

Comments

I think there is a three papers used in opencv mser :

  1. Robust Wide Baseline Stereo from. Maximally Stable Extremal Regions. J. Matas. 1,2. , O. Chum. 1. , M. Urban. 1. , T. Pajdla.
  2. Linear Time Maximally Stable Extremal Regions. David Nistér and Henrik
  3. Maximally Stable Colour Regions for Recognition and Matching. Per-Erik Forssén

paper 3 is for color image

LBerger gravatar imageLBerger ( 2016-03-25 15:10:40 -0500 )edit

Yeah I know, I have researched all of them.
The thing is I see unadequacy between research paper for color images and reserahc paper for color images. I was hoping someone know about source code for color images, which I can't find.
I can't demonstrate here any result to explain difference between OpenCV results and paper.

Davit gravatar imageDavit ( 2016-03-26 04:44:03 -0500 )edit
sai kethamakka gravatar imagesai kethamakka ( 2016-03-27 20:08:05 -0500 )edit