Thermal image, find fault

Hi all,

I have a thermal image like below. I pasted link there.

I want to find a fault position over there, but image have simultaneous places like lines. Can we bypass this, because when we convert it gray scale and threshold after that the fault place and normal place is being together.

Can we remove this lines from image, and image can be improved?

thanks all.

This is the pseudocolored rgb image:

this is binary and image processed.

Answer 1

[Image Analyst]

Without the colormap, this image is almost worthless - it's impossible to get back the original thermal image. Why don't you have the original image? It starts out as a monochrome image and then you apply a colormap and save it. Why can't you just not apply the colormap? You have the image, just save it without any colormap. Analysis will be a lot easier after that.

[ali]

this image is taken from video. It is one of the video's frame. The orinal one is like that.

the original; http://img476.yukle.tc/images/6170infrarouge_2013-03-01_20-34-23.jpg

and want to find this fault; http://img476.yukle.tc/images/64151.jpg

[Image Analyst]

The first image can't be displayed.

I've used thermal cameras not too long ago. I know that you can specify different colormaps - I've done it myself. Specify a gray scale color map and this will be possible and won't be too hard.

If you insist on using an unknown colormap then about the best you can hope for is to get the mean RGB for each column and look for anomalies as you move across columns, and that is only if you have vertical "rips" in your image like your second image indicated.

[ali]

You suggest when we take real video from camera, the format should be grayscale?

Exactly the mean RGB for each column what does that mean colud you please more obviously with an example.

Thanks so much for your great attention.

[Image Analyst]

Yes. I'm almost certain (not 100% certain because I haven't used all thermal cameras out there) that you can get a grayscale image out. You can use the mean() function to get the mean intensity going down columns.

[ali]

thanks for attention We have to use mean function for all of columns for example the image shoud be 512*512.