Hello Dear Friend Sasaki,

Whether you use a SLR or a DSLR, in my opinion, it is not going to help you in your temperature imaging application! Has it been that easy, people would not have made the Thermal Imaging Cameras for the accurate measurement of temperatures.

Therefore I would suggest you to use such Thermal Imaging Camera for your application. Off-course these are very costly and the pixel resolution is low, yet they give images in say 640x480 VGA size resolution through extrapolation.

The problem with the type of camera you are using is that the lenses/optics are primarily designed for the visible light and many good quality camera have lenses coated with IR filters. Kindly note that thermal imaging requires IR (infra-red) radiations and it will not work on your camera. Even in low cost cameras, if the lenses are not coated, a thin IR filter film is placed above the CCD/image sensor to block the IR radiations hence that too will not be of use for you. Unless you dig inside your camera, and do the R & D (Repairs and Damages!!!!) you might not get anything at all!

Now, coming back to your image processing! Your algorithm and logic might be correct and working perfectly alright but since your camera is not producing the expected linear change in your image, you are getting false or wrong results.

Therefore ensure first that your camera is giving you the desired results or not and then attack on your image processing algos if needed!

Hope my explanation is helpful to you.

Thanks and best luck!

Dear Prasanna, Sasaki is not trying to measure the temperature directly, relying on the IR sensitivity of the camera, but using temperature sensitive paint (TSP, it would have helped to include this acronym :-).

I have no experience with such paint whatsoever, but you appear to have a ranging problem: either the temperature distribution is out of range for the paint, or the light distribution is out of range for the camera. The first one you can check in the data sheet of the paint, probably. The second one you can check easily by looking at the histogram of the picture on the camera itself. Hope this helps!