Another smartphone thermal imager from i3

[Nash]

Thank you. I will ask them and see joe-c software

[Ben321]

Recently came across this - surprised nobody found it earlier.

http://www.i3-thermalexpert.com/

"Thermal Expert" by a Korean thermal imaging company "i3" (not to be confused with L3, who also make thermal imaging gear).
Details are vague but this is what we know;

• Release Q3 this year
• 384 x 288
• Android only at launch
• -20 to 100 degrees Celsius

Images supplied by i3 on their website appear to be very good quality - on par with Therm-app? Curiously the GUI has 2 video buttons.




Obviously the big question is price - I can't find any mention of price except here: http://www.tempsensornews.com/thermal_imaging/prototype-low-cost-smartphone-thermal-imager/

Which states "Numbers like $500 per unit with more than 3 times the 32 K pixels of the ~$250 Seek Thermal and were heard resonating on the Expo floor."

Also here: http://electronics-eetimes.com/images/01-edit-photos-uploads/2015/2015-06-june/2015-06-19-eete-jh-yole2.jpg

Which states "<$500 USD" but also a 2016 release date which is not in line with the "Q3 2015" figure on their webite.

EDIT: They have a video on their youtube channel:

It looks like the framerate isn't the best, and the lens has some sharpness issues at the edges, but still, it will certainly be the best option available at it's price point (assuming the $500 figure is true)

Not sure where you got your $500 price from, but last time I contacted them and asked (last year) for the cheapest unit with the cheapest lens, they said it was over $1000. Did they suddenly just double the price like that? Or are you remembering the info wrong? Where did you even get your $500 number from? I can't find any prices on the website. I had to contact them to get the price.

[frenky]

This is an old mail I got from them... I have then bought TE Q1+ for 599$.

[Ben321]

This is an old mail I got from them... I have then bought TE Q1+ for 599$.

Wow! Those guys really jacked up the price recently then. The cheapest one I saw on the prices they showed me was like $999. Why would they do that? They are just cutting out potential customers, including me.

[CatalinaWOW]

One of two reasons.  They are selling all they can make at the higher price.  Or their cost of manufacture doesn't let the lower price work.  Bolometer yield could be low, or any number of other things could raise the costs beyond what they projected.

[brammeijer]

im thinking of upgrading my seek
i was thinking of this or a therm-app and these went from 499 to about 999 for some reason  (therm-app also 999)
what would be the better buy, how does the Q1 Plus compare to a therm-app with 19mm lens

[DaneLaw]

Wauue thx guys very informative thread and certainly interesting 110k alternative to Opgals Thermapp..
sadly also been hooked by this expensive IR world syndrome in the seek for details and been trying to comprehend an overview from a noob perspective, as these tool are insanely pricey here in Scandinavia Denmark.and the prices are useally inflated by a factor of +80 to 90% in regards to USA retail after DK techtaxing, import and further +25% VAT..
I have looked at the Seek pro, and most sellers here list it at 800 to 850USD.

Finally found a Seek pro unit so-called new to 290USD plus a minor shipping-fee from US and took the plunge and label it for for no further taxing as the limit here is insamely low 13USD, sadly this socaled new unit Seek pro was IOS so reached out to Seek to ask if I could just use adapter with datapins' as in lightning to Micro USB.. they where not 100% sure, as the hardware was not identical, but mostly the other way around as Apple got some hardware surface mount attachment to make it valid but in theory it should work.
In the end' didn't wanna risk it' as 290 USD is still some dope and not least it wouldn't be fair in regards to the seller I was purchasing it from it from' to experience and not check it firsthand on a native system if there was any problem.
So instead jumped on my first IOS device,. and picked up a new 292 Euro IPad 2018 vers. with a lightning port from the UK and been having a field day with this IR-world. (very fascinating)
Here is some examples (link below) on what Im getting with this device,but the focus distance maxes out at 20cm, instead purchased & for at the moment got some cheap IR macro-lenses in transit. 1x20mm to 4½$ and 1x12mm 8$ incl shipping if they any good, don't know, I'm a noob and mostly learning as I go..but coming around this forum will be a helping hand and was not familiar that EEVblog was active in the world of IR cameras, but there is certainly some competent people in here, who knows their stuff.

Below videos all with Seek Pro IOS. (el cheapo-device at 290 USD with a 240x320 sensor and sadly the trees doesn't grow into the sky at this low price ballpark and after reading this thread got an appetite for Thermal Expert low key editions, if they were valid around the  500 USD ballpark. but if the price has raised here 2 years later with a factor of+100% (+999USD) it goes without saying it tumbles the purchase, - as after mandatory danish taxing and another +25% VAT it will not make sense. particularly not in regards to the result below from an 290 USD value..)
https://www.youtube.com/channel/UCTJtb7oyUVo0CVfd0RQ5rsw?

Mediainfo dump Seek pro, avarage-FPS is listed at just below 17fps, but the manual shutter alignement, does dip FPS when in play.

Complete name : C:\x\PIR-206_4.mp4
Format : MPEG-4
Format profile : Base Media / Version 2
Codec ID : mp42 (mp41/mp42/isom)
File size : 8.46 MiB
Duration : 1mn 14s
Overall bit rate : 947 Kbps
Movie_More : Taken with iPad7,5,iOS 11.3
Performer : SeekwareBlendModeNone
Encoded date : UTC 2018-05-08 08:52:35
Tagged date : UTC 2018-05-08 08:52:37

Video
ID : 2
Format : AVC
Format/Info : Advanced Video Codec
Format profile : High@L2.1
Format settings, CABAC : Yes
Format settings, ReFrames : 2 frames
Codec ID : avc1
Codec ID/Info : Advanced Video Coding
Duration : 1mn 14s
Bit rate : 881 Kbps
Width : 480 pixels
Height : 360 pixels
Display aspect ratio : 4:3
Frame rate mode : Variable
Frame rate : 16.824 fps
Minimum frame rate : 1.685 fps
Maximum frame rate : 40.000 fps
Color space : YUV
Chroma subsampling : 4:2:0
Bit depth : 8 bits
Scan type : Progressive
Bits/(Pixel*Frame) : 0.303
Stream size : 7.85 MiB (93%)
Title : Core Media Video
Encoded date : UTC 2018-05-08 08:52:35
Tagged date : UTC 2018-05-08 08:52:37
Color range : Limited
Color primaries : BT.709
Transfer characteristics : BT.709
Matrix coefficients : BT.709

Audio
ID : 1
Format : AAC
Format/Info : Advanced Audio Codec
Format profile : LC
Codec ID : 40
Duration : 1mn 14s
Source duration : 1mn 14s
Bit rate mode : Constant
Bit rate : 64.0 Kbps
Channel(s) : 2 channels
Channel(s)_Original : 1 channel
Channel positions : Front: C
Sampling rate : 44.1 KHz
Frame rate : 43.066 fps (1024 spf)
Compression mode : Lossy
Stream size : 578 KiB (7%)
Source stream size : 579 KiB (7%)
Title : Core Media Audio
Encoded date : UTC 2018-05-08 08:52:35
Tagged date : UTC 2018-05-08 08:52:37

But so far I do cherish the above thermal- result and I reckon I need to put out at least 3 to 4x times that to get any obvious better result at 320x240 or above sensorspecs, are there any obvious choices in the budget sub 1k USD ballpark. if one wanna upgrade from Seek pro, other then Therm-app 9hz vers. or Thermal Expert or a hacked Flir E4. which all seem to start from around 1000 USD.?

-new to the forum will take a look after a Seek pro thread and ask if somebody perhaps have tested their IOS version on an Android edition true basic adapters, but if its common knowledge, please state it..

[wraper]

Wow! Those guys really jacked up the price recently then. The cheapest one I saw on the prices they showed me was like $999. Why would they do that? They are just cutting out potential customers, including me.

Did you get a quote directly from them? They got distributors in some countries who want twice of original price.

[mahony]

Yes, I would definately contact I3 directly - as far as I know the prices are still 499$ for TE-Q1 and 599$ for the TE-Q1+ plus shipping fee of 80$ plus VAT/customs of your country.
Actually there is a distributer in Austria that sells the cameras on ebay and an online-shop (719€ for the TE-Q1 and 899€ for the TE-Q1+ currently). Link: http://highvoltageshop.com/epages/b73088c0-9f9a-4230-9ffc-4fd5c619abc4.sf/de_DE/?ObjectPath=/Shops/b73088c0-9f9a-4230-9ffc-4fd5c619abc4/Categories/Waermebildkamera

I got my TE-Q1+ directly from Korea for ~800€ (incl. shipping and 19%german VAT/customs) but at ~1.05USD/€ around january 2017. At the moment with exchange rates around 1.2USD/€ I would should get away with ~700€ delivered for the Plus-version and slightly below 600€ for the standard TE-Q1. So there is quite some offset compared to a purchase of the re-sellers...

[wraper]

Yes, I would definately contact I3 directly - as far as I know the prices are still 499$ for TE-Q1 and 599$ for the TE-Q1+ plus shipping fee of 80$ plus VAT/customs of your country.

The issue is they'll forward you to distributor if there is any in your country.

[DaneLaw]

Any here made any inquiries about IOS support (hence the phrase "comming soon" Im not familiar with how long that has been stated on there site)?

■ Compatibility : Android (iOS will be soon)

[Klaus]

I just installed the new android app, some new things, like better color off the min and max temperature and some image enhancements.

[frenky]

Anybody has any idea why there are such pixel "steps" on this image taken with TE Q1?



Upscaled:

[bap2703]

It looks like it was first upscaled with a 'nearest neighbor' type algorithm, and only then a real interpolating algorithm was used.
Leading to big blocks with a smooth but tiny transition between them.

[frenky]

It's weird because there is no need for upscale since T.E. has 384x288 sensor. 
I'll redo capture with TE conencted to PC...

[frenky]

Ok now I feel really stupid... 

But hey in the name of science I shall publish my "investigation".

I took original 384x288 image upscaled it to 400% drawn two lines on it (dark gray and white) and downscaled it again to 384x288.


As you can see I got very similar artefacts with pixel steps.
Visible steps are because of low resolution of sensor and low angle of heater relative to horizontal lines.


(enlarged image so you can see pixels better)

[Ultrapurple]

Interesting.

I just ran a test with a similar-enough heater using a 640x480 ThermApp Pro (the only camera I had to hand). Although the resolution is higher than the TE I'd expect the results to have broadly comparable moiré. But I only managed to make it visible by playing around heavily with the palette. Here I've shown the original and the enhanced version. The B&W image is as it came from the camera, the psychedelic one started out in 'rainbow' palette and then had all the dials turned up to 11.

Finally, I took an Ironbow image and (a) shrunk it to 25% then re-sized it to the original and, separately, enlarged it by 400% then shrank it back to the original size. There is a perceptible loss of quality in the first but none in the second. All manipulations were done in Paint Shop Pro X.

It looks to me as though there's something iffy about the way the TE produces its JPGs. Some may remember I had problems with the initial driver software for my ThermApp Pro, where it was only using 384x288 pixels of the 640x480 sensor. I wonder if there may be something odd like that going on with the TE?

[Vipitis]

I suspect it's a result of post processing. Moiré possibly, but unlikely as it shows up on CMOS sensors with low resolution due to debayering of the bayermatrix and interpolation of subpixels for RBG values. That's way you see green or purple in some grey striped shirts for example. Remember that upscaling to a non multiple resolution also counts as moiré.

Aliasing would a term to use. If you upscale or downscale you will make jumps. No doubt.

I also seen artifacts with sharpening applied. Or when I open and save a radiometric jpg through FLIR Tools mobile and compare that to the direct output of the MyFLIR app. It's really difficult to compare for me as my capturing app only allows for a single file to be saved and I would need to do up to eight individual captures to match all modes  - which is a hassle due to the big with settings in ThermalCamera+

Remember that artifacts of post processing follow a pseudo log curve in their impact vs resolution. The lower you initial resolution is the more notable the the effects.
 Modern digital cameras and even phones these days are doing a whole lot of computational imaging because their input a lot of data with 20MP images or 2x5.2k video. Thermal cameras don't nearly get close to those amounts of data. Therefore I see huge potential in a kind of computational MSX that used image recognition from a common database and rearranged the low amount of thermal pixels to every single part of the visual images and increase edge representation by this. Maybe just a thought I should patent ASAP or the future.

[frenky]

So I tried to capture similar image (metal fireplace) with PC runing original TE software:
Not sure why did they even bother writing such crappy software...



Joe's app on the other hand is working fantastic:

[Vipitis]

Looks to me like those are artifacts from sharpening. Even though your scene is very high contrast there seem to be some noise at the pixel level(or an uneven) surface. I am guess that the sharpening is applied after the histogram application and therefore the tiny changes are greatly amplified.

On the other hand a lot of upscaling algorithms add a similar kind of sharpening to the image. But it's not working for his kind of inputs as seen(its way too strong). I have had this issue where due to the upscaling and sharpening all corners between high and low temperatures had a rainbow line between them because the algorithm made up new pixels in between by interpolation the high and low value of adjacent pixels. And came up with vastly different inbetween values. The results don't look great and are only getting worse by additional sharpening.
This is more prelevant in palettes that run through multiple colors and images that have a hard border with a larger temperature drop/rise.


There is a whole process called deconvolution that can be used to increase resolution, but it's a very difficult problem and not applicable for real time imaging.

[Ultrapurple]

X-ray imaging has been around for quite a while - see, for example,

     

Click any image to embiggen and learn details of the setup used, or click here for more

[frenky]

Combined x-ray & thermal images are really cool. 

[frenky]

I have found a way to disable high contrast pixelation effect on T.E. android app. It's "Enhance image" setting in the menu.
While this option does improve low contrast scenes it also ruins high contrast scenes.

With "Enhance image" enabled (and automaticaly upscaled and saved to JPEG in the app):


With "Enhance image" disabled (and automaticaly upscaled and saved to JPEG in the app):


Same image converted from csv to png in my ThexConvertGUI app and upscaled with waifu2x:


I have decompiled apk from android app to find out why images have so strong jpg artefacts. Well JPEG compression is hard coded to 90. 

Code: [Select]

source.compress(CompressFormat.JPEG, 90, outputStream2);

[Ultrapurple]

Wow! That is a substantial improvement in quality. Kudos for accessing the data and determining that the poor image quality is simply down to an incorrect setting on the compression algorithm. Perhaps they meant to 'keep 90% of the quality', rather than 'throw 90% away'.

JPG's effect on straight lines is one of my pet hates. I work in technical publishing and so many people seem to think JPG is the right format to save everything - yet for things like screendumps and other line art its artifacts irretrievably knacker the image. JPG has its place, but high-compression JPG can be really, really nasty - even on photos. On text it's a disaster:



(and yes I know that image is a PNG - it was deliberately saved that way so that the 'uncompressed' part remained uncompressed).

[frenky]

There is nothing worse than jpg for graphics with lines or text. People who use it for anything other than photos will burn in digital hell.