Extraction of ROI

[shangar]

Hello friends

             for my project I need to macth the specific ROI and extract it. Then I want to compare it with some Predefined pattern.. Now I design a code for pattern matching in labview using vision express .. Any can please help me to how to extract my matched image and comparing it also

                advance thanks

[ZcuBa]

what do you mean by "match the specific ROI" ?

a Region Of Interest is by definition a specified region, that is somehow provided by something/someone.
to match it, simply read it, and use the data.
it can be done by unbundle by name, and take the part of the region definition that you want.

If you mean that you manually "draw" a ROI on an image controller, and wants to get the information, right click on the terminal, and create the property node "ROI"

If you have some sort of implicit ROI in the image data, you need to identify the data, using either edge detection or pattern match, and then hope that the desired data is located, and then use that to generate a ROI specification.

perhaps you are really trying to ask for a countour analysis ?

to compare it with some predefined pattern, you should consider how to express the predefined pattern.

do you want to find a pattern in an image ? or do you need to know the polygon that the user have drawn ?

extracting an image is done by the image extraction vi, it takes a ROI to specify what to extract..
comparing can be done in many ways, do you need to compare the intensity, the patterns, the colors ?

[shangar]

i want to compare the pattern of my ROI image... in that i have to compare it with two or more images .. for that where i have to maintain data base...?

if u have any example codes please send me regarding comparision with two or more image...

[ZcuBa]

You need to think about how you want to compare images, and what kind of result you are expecting from a comparison.
that depends a lot on what the application is.

you say you have a data-base to maintain, what is the purpose of that?

Do you need to know about the color information of the two images ? (or is a gray image sufficient ?)
Can you live with only comparing the color information ? (is this label red or green etc. )

Is the pattern in the image relevant, and if so, is the specimen always found at the same position (distance and traversed) and angles relative to the camera, or do you need to scale and rotate images ?

Does the perspective change of the images from frame to frame ?
Do you need camera ego-motion estimation?
Can the subjects compared be partially occluded?

I would suggest that you start by looking at the shipped examples, to see what NI engineers have proposed of solutions to well defined problems.
-> help -->Find examples...

for a pattern comparison, you could try using
IMAQ Setup Learn Pattern 2 VI
and
IMAQ Learn Pattern 2

to create the database of learned reference images.

and then later, where you want to actually perform the comparison, for all learned images do:

IMAQ Setup Match Pattern 2
and
IMAQ Match Pattern 3

this will find the number of likely occurrences of a pattern seen in a learned image, within a test image.
the result is a number of matches, each containing the location of the learned reference image.

[CarlosOtiniano]

In the following we proceed to function block search pattern extracted in the previous process (the parameters as rotation angle and minimum score is inserted into SETTINGS control), extract the output of the search function to get the position values indicators that will be displayed on the front panel)