CODEWITHSUNDEEP
imageimage-processingimage-segmentation
wanglinski
wanglinski

Reputation: 112

image segmentation for grains

I am trying to separate the different kinds of grains in an image. And sometimes the image also contains some impurity substance which need to be considered as an extra type. here are some example images:

corn and beans img

long rice and wheat img

I tried to find a general method for the different pics, but the result is not good enough. I used flood-fill and some gradient method to get the regions, and try to use clustering method to classify the contains, but the feature selection is a hard problem, I try gabor filter, but it cannot get me a clear boundary, and so does the classification method such as kmeans.

Any ideas about segmentation, getting the contours or classification will be appreciated. thanks!

I try to post some more pics of my current results, but I am sorry that there is the 2 pics restriction for the beginner here.

Upvotes: 0

Views: 565

Answers (1)

Victor Oliveira Antonino
Victor Oliveira Antonino

Reputation: 988

It's almost a craft work dealing with image processing problems. I would suggest you to use a robust library (such as OpenCV of course) and use cvFindContours function to identify the contours. Also, search for mathematical morphology. Basic operators such as erosion and dilation may help you since areas of foreground pixels shrink in size, and holes within those areas become larger and vice-versa. Working with color segmentation is also helpful but you might have some troubles since grain color is not uniform. Lastly, feature extraction is another way out. Scale-invariant feature transform can be used to identify every single grain on the image, based on the fact that it is invariant to linear transformations and illumination issues. Hope it helps.

Upvotes: 1

Related Questions