What are the rough steps in obtaining and processing an image from an infrared camera?

Question

I have asked the same question in stackoverflow, but since I am not getting any responses , I am going to ask it again here.

My project is to design an interface between an IR camera (Xenics XTM-640) and an ARM/FPGA unit and then process the data to make some comprehensible image.

At the moment I am working on creating timing waveforms to control sensor and capture of frames and I am planning on how to go on from there onwards.

From what I can see there is number operations which can be performed to make the image clearer. Terms I have came about such as:

• quantisation
• convolution
• Gaussian filterring
• contrast enhancment
• noise cleaning/noise smoothing
• neighbourhood operation
• normalisation
• edge crispening

What is the difference between preprocessing and processing?
Is there a preferred order to performing this steps?
Are some absolutely necessary, while others not so much and can be skipped?
How would this differ between processing normal light vs IR radiation?

Ideally I've been looking for a flowchart/diagram that shows all (or some) of the stages from capturing IR radiation to displaying false colour image on a screen.

In the "Digital Image Processing" by William K. Pratt it seems the chapters are broken down into

1. CONTINUOUS IMAGE CHARACTERIZATION
2. DIGITAL IMAGE CHARACTERIZATION
3. DISCRETE TWO-DIMENSIONAL LINEAR PROCESSING
4. IMAGE IMPROVEMENT
5. IMAGE ANALYSIS

Is that roughly the correct chronologically description of the whole process?

Answer 1

It's hard to give a lot of useful information without knowing the details of the signals you are getting from your hardware. Typically the output from any camera is an image, so you would need to decide what is wrong with the image before trying to come up with image processing techniques to solve that problem, you don't seem to have a clear problem specification at the moment.

What is the difference between preprocessing and processing?

Semantics. It doesn't refer to anything specific in the signal processing domain.

Is there a preferred order to performing this steps?

Yes but you need to decide which are necessary before worrying about the order.

Are some absolutely necessary, while others not so much and can be skipped?

It depends entirely on your source data. If your IR camera outputs a clear crisp image that contains all of the information you need then none of the steps will be necessary. If the image is completely washed out then you will need to do contrast enhancement, otherwise it can be skipped. Really nothing useful can be said without more information about your data.

How would this differ between processing normal light vs IR radiation?

The usual caveat applies that it depends on your data but in principle there is no difference, you are just capturing a different part of the light spectrum.

Is that roughly the correct chronologically description of the whole process?

No not really, the process will be informed by your source data and your desired output. It's unlikely that all of those topics will apply, more likely one or two of them will and not necessarily in any particular order.

If you want more concrete advice I suggest you explain your procedure in more depth, with the specifics of the data you are acquiring and the results you expect.

Answer 2

Like PAK-9 mentioned it is little hard to tel about the processing required without knowing about the data form the camera. These answers may provide some info to you.

Are some absolutely necessary, while others not so much and can be skipped?

Not all blocks are necessary.I don't understand what exactly you mean by convolution and neighborhood operation. My choice of processing will be Noise filtering,Gamma correction(Depending on the sensor data),Edge sharpening and Contrast enhancement.

Is there a preferred order to performing this steps?

Not every processing has to be done in ordered way. But its good to have sharpening after noise filter and Contrast enhancement after gamma correction.

How would this differ between processing normal light vs IR radiation

The difference between processing normal light and IR light is if you want a colored image your sensor will output RAW images in Bayer format(mostly). You need to do demosaicing to get the colored image.In case of monochrome it won't be necessary. The IR image sensors behaves like a monochrome sensor.