artificial image BLURRING in the Flir Ex series?

[calel]

apparently no other topic mentions it (except once) so I make a dedicated topic just for that one

so this isn't about the noise generator which everyone knows exists in the Ex (and Exx too apparently) only about the blurring


could it be that they also added some blur for the Ex?

evidence:

Ex image (320x240)



Exx image (320x240)



more then evidence, this is proof: clearly this isn't the noise generator at work (I think both cams were hacked to turn it off anyway) so the difference between Ex & Exx at same resolution can only be explained by the presence of some added blur on the Ex right?  (case in point I can take the Exx pic, add in blurring effect and I'll have the same pic as the Ex one, with only a difference in brightness/contrast)

else what other explanation??

[Pinkus]

The Exx comes with a much larger lens which also allows (needs) to focus. The Ex only has a small fix focus lens. Similar to 'real' cameras: the ones with a small fix focus lens are related to worse picture quality compared to a DLSR.
Side note: Because of the superb image details of the Exx, MSX usually is not needed here (though there are a few situations where it can be helpful).

[Bill W]

The Ex image just looks focussed too close, while a true added blur should be equal at all distances.

Compare the CPU fan and top right of the image (fairly similar, maybe only the cheap lens making a difference) to the deeper parts in the mid right

[calel]

bummer I forgot about the whole lens thing...again

but why's the Ex image so much brighter

[calel]

btw when you say cheap lens do you also mean the material used?
in another thread Frazer suggested maybe the Ex lens made of some Zinc compound whereas Exx lens made of Germanium anyone confirm the Ex & Exx lens dont have same ingredients?

[Fraser]

To clarify, I suggested you research Germanium, Chalcogenide IR Glass and ZnSe. All three materials are usable in LWIR thermal imaging camera lenses. I did not suggest that the Ex uses a ZnSe lens, as that is most definitely not the case.

The Ex series uses a molded Chalcogenide IR Glass lens
The Exx series uses a diamond lathe cut mono crystal Germanium lens material

The Exx lens is superior to that of the Ex in all respects except, maybe, use at elevated ambient temperatures.
Diamond cut Germanium lenses are far more expensive to manufacture than molded Chalcogenide IR glass. As a result, you will not normally find Germanium lenses in lower cost thermal imaging devices. Two of the most expensive parts of a Microbolometer based thermal imaging camera used to be the sensor array and the Germanium lens. Chalcogenide IR glass has massively reduced the cost of the lenses and miniaturisation of microbolometers has reduced their cost significantly as well.

[Bill W]

Cheap as in if FLIR paid more for the lens it could be better.  Material costs are less relevant as you can make a very good lens just from ZnSe, or pure Ge, or silicon or even Ge-chalcogenide if prepared to pay for added processes, lower yields, extra elements or other complications of doing a good job.
The best lenses can often include several materials to optimise the design

Or you can do 'good enough to sell' for low end models and hold back the better performance (& higher costs) for the next model up which you charge for (in) appropriately.

I have just run the E4/E8 lens data, and it is interesting given the images you posted above.
45° FoV on a 320 x 240 @17µm is a 6.6mm focal length
FLIR state f/1.5
The near point of the camera is stated as 0.5m
That combination says best focus at 0.7m with a far point at 1.2m
Or to keep infinity in good focus you set for 1.7m, and the near point is 0.85m.

These are the problems inflicted by fixed focus, or 'focus free' as the spin doctors in the FLIR lying department have called it.

[Fraser]

Bill,

Do not get me started on the Marketing Department descriptions of focus systems. They push the positives and forget to mention the negatives of course ! Fixed focus can be very convenient, especially in scenarios where the user cannot be expected to keep adjusting lens focus, as you well know  For general thermal imaging the E4 lens does its job reasonably well but the superior quality manual focus of the Exx series does show what can be achieved in terms of improved image sharpness. Those used to using cameras with fixed focus lenses can find the manual focus F1.0 lenses challenging due to the much reduced DOF and need to adjust the focus often when observing targets at different distances. But then nothing comes for free in life 

I have seen thermal cameras advertised as "Auto focus" when in fact they use a fixed focus lens. The sellers claim that the huge depth of field means they are "Auto Focus" in terms of providing a reasonably focused image at many distances 

Now true motorised auto focus on a thermal camera is nice to have.....BUT..... the auto focus systems I have on my cameras are like early SLR and Camcorder auto focus systems..... not always that good at picking the exact sweet spot for best image focus. The small DOF on professional thermal cameras with F1.0 lenses does not help matters either.

All good fun 

Fraser

[Bill W]

The Exx series uses a diamond lathe cut mono crystal Germanium lens material

That is one reason the Exx has manual focus then, as germanium changes refractive index over temperature quite significantly.
Being f/1.0 will not help either as that reduces depth of field.
All what you expect to do to get good images.

Following might be of interest from Lightpath (who make the moulded lenses for several different camera suppliers)

https://www.lightpath.com/product/thermal-imaging-assemblies/
https://www.lightpath.com/wp-content/uploads/2019/09/Thermal-Images-from-LightPath-Lenses-on-BAE-Detector.pdf

Reasonably technical set of papers explaining lens designs that are now openly available if somewhat dated:
https://www.lightpath.com/technical/
under 'Infrared Optics'.

My figures suggest FLIR will be struggling to make the Ex work over a reasonable set of distances and, if that 0.5m is correct, are making an odd choice in presetting the focus.  With fire cameras we had a much tighter set of likely ranges to deal with, and generally got a 1.5m near point and almost making infinity.

If you mean 'auto focus' on ISG fire cameras, that was a *completely* different thing to be explained another time.

Bill

[Vipitis]

FLIR does use blurring. As an upscaling method for their Lepton based Products. I know from the S60 and S62 I own (Lepton 2.5 and 3.5) that they upscale the thermal data to 4x the resolution (in two axis = 16x the pixels). Radiometric JEPGs shot with a 80x60 camera read as 320x240 and 160x120 pixels are pushed to VGA. This causes the image to look badly blurry and it's fairly unusable. But this also hides noise and out of focus elements. As my new phone does no longer work with the 3rd party app of my choice and the developer confirmed me that the app is basically shelved. I am looking into a way to extract the "raw" resolution image from a blurrily upscaled radiometric JPEG again.

[bap2703]

This is not "blurring", this is interpolation.
https://matplotlib.org/3.1.1/gallery/images_contours_and_fields/interpolation_methods.html

But yes for very low resolutions the brain might prefers a sharp low-sampled image instead of an highly interpolated version.

[Bud]

@OP post the original images. What you posted were not the original images, they were post-processed ones in Flir software.

[Bud]

FLIR does use blurring. As an upscaling method for their Lepton based Products. I know from the S60 and S62 I own (Lepton 2.5 and 3.5) that they upscale the thermal data to 4x the resolution (in two axis = 16x the pixels). Radiometric JEPGs shot with a 80x60 camera read as 320x240 and 160x120 pixels are pushed to VGA. This causes the image to look badly blurry and it's fairly unusable. But this also hides noise and out of focus elements. As my new phone does no longer work with the 3rd party app of my choice and the developer confirmed me that the app is basically shelved. I am looking into a way to extract the "raw" resolution image from a blurrily upscaled radiometric JPEG again.

Are you talking about raw thermal images or the visual jpegs when you say "Flir does use blurring"?  Flir must have some type of processing because of dead pixels in the sensor. I'd think raw images already have that processing applied. But the JPEGs are separate and may have their own post processing. For JPEGs you should first minimize the compression in the camera settings if there is one. A trick to improve visual pictures quality in the liberated Flir E4 is to toggle the measurement control, for instance switch the measurement cursor off and on. This will cause the camera to apply minimum JPEG compression setting, thus improving the image quality (at the expense of increasing image file size). You can only do this on the liberated E4, because it has measurements templates modified for that. There was no way to force low compression setting by default, you must toggle the measurement tool.

The scripts for extracting raw images from radiometric JPEG were poster in the monster E4 thread. I recall they were PHP scripts.

[Vipitis]

I will send you a radiometric from the S60 (Lepton 2.5) which is interpolated(blurrily upscaled) to QVGA. I don't want you to pay attention to the thumbnail, if you open it in FLIR Tools it will show as 320x240 and will give you this many measurements as well with odd decimals all over the place. (there is an embedded visible in two different scales as well).

there will be some captured with the official app (those also say 2mm lens) and some with the Thermal Camera+ app by George Friedrich (that show VGA resolution as well???).

I am aware of the scripts but I haven't managed to do it with any of my files - but also haven't tried in a while.

I might be able to send some S62 (Lepton 3.5) Radiometric JPEGs in the following days, I haven't looked at them in FLIR Tools myself yet.

[Bud]

This topic is about E4, not sure what i can do with S60 images 

[Bud]

I do not think you can extract the "raw resolution" raw thermal images from these files. It appears the raw thermal images are already stored upscaled in the radiometric JPGs. So the "raw resolution" information is lost.

[bap2703]

I do not think you can extract the "raw resolution" raw thermal images from these files. It appears the raw thermal images are already stored upscaled in the radiometric JPGs. So the "raw resolution" information is lost.

Not necessarily.
It's possible that the source data is just surrounded by interpolated points.

Even if the data is not just interpolated but completely resampled it could be possible to reverse the operation.
But for that the resampling method and the interpolation method needs to be found.
That can be done but is probably a lot of work with regard to the expected results.
Just resampling at the original resolution won't do?

[Bud]

If you want to take a stab at it here is the raw binary files which can be opened in ImageJ in 16-bit big endian format, 240x320 and 480x640.