Sparse data -> "temperature" map

asked 2013-11-28 08:10:22 -0500

ainawing gravatar image


This is less of a technical problem and more of an image processing question.

I'm using the Java library provided by opencv.

I have a sparse number of data values. They all have a position on a "map" (i.e. Picture). They are also connected in a graph-like manner. Currently I'm generating a gaussian mask, scaling it to value and adding it to a 1-channel Mat in the correct place, while only taking the maximum value of the roi and gaussian mask. Then I apply a colormap and transfer it into a 3-channel Mat and there you are. (here's a result of the current process)

However, this doesn't look very good.

Now, what I'd like to get as a result is a more coherent kind of temperature map. The point values propagate through the connections and I'd like to reflect that. Getting a gaussian overlay over the connections that get linearly adapted as it goes from start to end. My question is, whether or not this can be done in any efficient manner in OpenCV.

I already thought about using delaunay and contour lines to denote the different color levels, but I can't think of anything that will make the processing even borderline efficient.

I'm a programmer, but not really apt in image processing. I hope someone here can help me. Maybe just point me in the right direction...



edit retag flag offensive close merge delete


I just thought about using a step by step interpolation of the data points, going through each connection. This still leaves a problem. How to do that and get the desired result... I don't see OpenCVs "Geometric Image Transformations" working that way. Is there any way to do a surface interpolation in OpenCV?

ainawing gravatar imageainawing ( 2013-12-02 06:30:39 -0500 )edit