Home Image-Pro General Discussions

Count crystals


I am using image pro premier and I am trying to measure crystals in the picture. Since some crystals are attached to each other, it is hard to measure the diameter of each crystal. I have tried to use "smart segmantation" to measure the crystals and used "split" to separate each crystal; however, the program does not count individual crystal accurately.  Is there any method that I can solve this problem? 

Thank you,

Best regards,


  • Hi Juhee,

    I am not sure what is the definition of "crystals", but if that's bright small particles, then you can use Band-Pass filter with the parameters as on my screenshot and Count with Auto-Bright (I also use Auto-Split with Boundary Shape option). The result looks like this:


  • 2020-06-01-185601

    Juhee --

    Yuri has provided some information about analyzing the crystals in your image.

    Below is another method to tackle your challenge.

    I this method the image is broken into TILES and the TILES are classified as
    • NEG (RED)
      -- No crystals present
    • LOW (ORANGE)
      -- Thin (dark) crystals present
    • MED (YELLOW)
      -- Intermediate crystals present
    • HIGH (GREEN)
      -- Thick (bright) crystals present
    The NUMBER OF TILES and PERCENT OF TILES in each CLASS are shown.

    If you are looking at your crystals over time this may be a helpful way to look at your samples.

    The TILE COUNT (DENSITY) can be increased by reducing the TILE AREA.

    I hope this information is helpful.

    -- Matt

  • edited October 2020
    I was wondering if you have a more detailed text for the steps you need to do to obtain the results. Is there a way we can know for sure that the object select are crystals?

    Thank you!

  • 2020-10-07-093805

    E.W. --

    The RESULTS that are extracted and reported are based on a LEARNING CLASSIFICATION in IMAGE-PRO.  Since the IMAGE-PRO CLASSIFIER is only "seeing" the PIXEL DATA, there is now way to "know for sure" that an area is primarily CRYSTALS.  I did a TRAINING for the CLASSIFIER and it picked the REGION MEASUREMENTS that is saw as consistent with my training.

    Imagine that you were attempting to classify ADULT HUMANS by WEIGHT into GENDERS.  You would probably have pretty good accuracy saying that ADULT HUMANS under 200 pounds are FEMALE and ADULT HUMANS over 200 pounds are MALE.  It would not be perfect but I am sure it would be much better than RANDOM.  If you add HEIGHT, you would probably become more accurate but still not perfect.  If you add HAIR LENGTH, you would probably improve accuracy but there are certainly HEAVY, TALL, SHORT HAIRED WOMEN.

    I hope this helps.

    -- Matt
  • Hi E.W.,

    First you have to define what is "crystal". Can you provide a screenshot with some drawing showing what you consider a single crystal? Is the segmented objects from my post from May 27 close to your definition of crystals?

Sign In or Register to comment.