enhancement: extract binary data from FLIR radiometric jpg

[tomas123]

suggest for a enhancement: decrypt binary data in the EXIF-Header from a FLIR radiometric jpg



FLIR Infrared Software stamps every picture with a flir-logo and doesnt support batch processing.

Newer Flir Thermographie Cameras saves the 16 Bit sensor raw datas in the Exif Header of a normal jpg.
With this 16 Bit raw datas you can generate a own false color pictures (imagemagick etc.) without flir logo.

exiftool can't decrypt the embedded raw values (binary)

exiftool.exe -v5 IR_0193.jpg
...snip
JPEG APP1 (42312 bytes):
    104e: 46 4c 49 52 00 01 00 00 46 46 46 00 00 00 00 00 [FLIR....FFF.....]  // magic bytes sequence / 8. Byte+1 = numbers of linked blocks (max 65532 Bytes see sample below)
    105e: 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 64 [...............d]
    106e: 00 00 00 40 00 00 00 0e 00 00 00 02 00 00 00 00 [...@............]
    107e: 00 00 00 00 00 00 00 00 00 00 00 00 00 00 a5 40 [...............@]
    108e: 00 00 00 00 6d df 65 d4 00 20 00 01 00 00 00 71 [....m.e.. .....q]
    109e: 00 00 00 01 00 00 02 00 00 00 09 ac 00 00 00 00 [................]  // start address 1. segment + length (address is relative to this block - see awk script)
    10ae: 00 00 00 00 db 1f ca ed 00 22 00 01 00 00 00 68 [.........".....h]
    10be: 00 00 00 01 00 00 0b ac 00 00 03 10 00 00 00 00 [................] // start address 2. segment + length
    10ce: 00 00 00 00 9b 9f 81 a2 00 21 00 01 00 00 01 05 [.........!......]
    10de: 00 00 00 01 00 00 0e bc 00 00 00 64 00 00 00 00 [...........d....] // start address 3. segment + length
    10ee: 00 00 00 00 6d 9d 1a a1 00 01 00 02 00 00 00 65 [....m..........e]
    10fe: 00 00 00 01 00 00 0f 20  00 00 96 20 00 00 00 00 [....... ... ....] // start address 4. segment + length (0x0f20+0x104e=0x1f6e)
    110e: 00 00 00 00 f9 d7 91 31 00 00 00 00 00 00 00 00 [.......1........]
    111e: 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 [................]
    112e: 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 [................]
...
    1f6e: 33 93 92 43 31 00 00 00 02 00 a0 00 78 00 00 00 [3..C1.......x...]  // magic bytes 0200 and sensor size in little endian 0x00a0 0x0078
    1f7e: 00 00 00 00 9f 00 00 00 77 00 00 00 00 00 00 00 [........w.......]
    1f8e: 00 00 00 00 00 00 00 00 f9 2e e7 2e fb 2e ef 2e [................]   // here starts the 16 Bit Stream
    1f9e: 05 2f f6 2e e6 2e f2 2e 01 2f f7 2e 07 2f ea 2e [./......./.../..]
    1fae: f4 2e fb 2e fa 2e 03 2f 01 2f 06 2f f1 2e 07 2f [......./././.../]
...
    Warning = Ignored APP1 segment length 42312 (unknown header)

With reverse engineering I decrypted the header.
In the sample above there are a header for 4 segments (4x32Byte Header)
The address of first segments stands (6*16+5)Bytes from begin of JPEG-APP1-block.
3-Byte-Start-Address:  00 02 00, then 3 Byte with the length of the segment 00 09 ac
.... Second segment starts: 00 0b ac / length: 00 03 10
....  Fourth segment starts 00 0f 20 / length: 00 96 20

The problem is that the raw sensor values are always written in different segments (from first to fourth).
If you open the camera-jpg with a flir software (Flir Tools, Quicktools etc.) the software rewrites this block and save the raw values on another position.

As workaround I look in the top of every segment for a magic byte sequence "00 20" followed by the sensor size in little endian (!)
here Flir E40: W/H = 160 /120px = 0x00A0 / 0x0078

With this information over the sensor size you can calculate the the required segment size for saving all sensor raw data:
Segment_Size= sensor_width x sensor_height x 2Byte(16Bit) + 32Byte(Header)

in this sample above from a flir e40 we need: 160px* 120px * 2Byte + 32Byte = 38432 Byte = 0x9620 Bytes
With this information we can check the start address of segments above and the length of every segment
-> we found that the 4. segment has a necessary segment length of "00 09 ac" bytes for saving raw datas
-> result: the sensor raw values saved in the 4. segment

In the 4. segment, after discussed header of 32 Byte (with sensor size etc.), starts the stream of 16 Bit Raw Values (little endian).

I wrote a short awk-script for decoding the raw values and generating a *.pgm 16 Bit picture.
This script only works for newer Flir Cams with embedded 16 Bit RAW Data. I tested it successful with pictures from Flir E40, Flir E30bx, Flir T400, Flir i60 and Flir P640 downloaded from wikipedia and flickr ;-)

Some older Flir Thermal Imaging Cameras save a PNG in the exif header (B50, B60, P60, i7).
For this Cameras you need another code (see sample http://www.nuage.ch/site/flir-i7-some-analysis ).

The working awk script generates a picture in portable graymap format (PGM)
http://en.wikipedia.org/wiki/Netpbm_format:

Code Select Expand

# cat  raw.txt
{
  for(i=1; i <=NF; i++)
  {
     # find MagicBytes off JPEG APP1 (Flir) [8. Byte = 00/01/02]
     # 46 4c 49 52 00 01 00 ?? 46 46 46 00 00 00 00 00 [FLIR....FFF.....]
     if ( $(i)=="46" && $(i)$(i+1)$(i+2)$(i+3)$(i+4)$(i+5)$(i+6)"00"$(i+8)$(i+9)$(i+10) == "464c495200010000464646" )
      {
         # search first 5 segments of JPEG APP1 (address header of segments start at 0x50, step 0x20)
          for (i1=5*16; i1<256; i1=i1+32)
         {
           # MagicBytes: Block_Size= sensor_width x sensor_height x 2Byte(16Bit) + 32Byte(Header)
           # Flir E40:  160x120px -> 120*160*2+20=38432 = 0x9620
             # Flir T400: 320x240px -> 00 40 01 f0 00 -> 0140h x 00f0h -> 0x025820 (153632 Byte)
           # calculate entry address of segment
           a=i+sprintf("%d","0x"$(i+i1+5)$(i+i1+6)$(i+i1+7))+8;
             # magic bytes for following sensor size
             if ( $(a)$(a+1) == "0200" )
           {
               # sensor witdh and height in Little-Endian
               w=sprintf("%d","0x"$(a+3)$(a+2));
               h=sprintf("%d","0x"$(a+5)$(a+4));
               #  calculatte required segment_size= width x height x 2Byte + 32Byte(Header)
               bs_sensor=2*w*h+32;
               # read real segments size from header
               bs_exif=1*sprintf("%d","0x"$(i+i1+8)$(i+i1+9)$(i+i1+10)$(i+i1+11));
               if ( bs_sensor ==  bs_exif )
               {
                    # skip header
                    a=a+32;
                    # !! here you can set own values !!
                    min=0;
                    max=65535;
                    # now write 16 Bit gray scale picture
                    print "P2", w, h ,max-min;
                    print "# this is a *.pgm picture";
                 print "# sensor size: ",w,"x",h;
                    print "# start address of 16 Bit raw data: "a-1,"d / first 2 bytes: 0x",$(a),"+ 0x100 * 0x",$(a+1);
                    print "# raw values from range from min ", min, "to max ",max-min;
                    for (k=0; k < w*h*2; k=k+2)
                    {
                         # max Exif segment size 65534 Byte -> go to next segment (add 12 Byte to start address)
                         if ((a+k-i+4)%65536 == 0 ) a=a+12;
                          tmp=1*sprintf("%d","0x"$(a+k+1)$(a+k))-min;
                         if ( tmp >= (max-min) )
                           print max-min;
                       else
                         { if ( tmp <= 0 )
                              print "0";
                           else
                              print tmp;
                         }
                    }
               }
           }
         }
      }
   }
}


I'm not using gnu awk with binary extension.
For native awk (bsd) I use hexdump for generating Asci-Code from the jpg-binary.
In this way I can feeding awk with ascii code without stop code 0x00.

Therefore you must use the awk-code in a hexdump pipe (alternate save hexdump-ascii).
Save the result as a pgm picture...

Code Select Expand

# hexdump -v -e '1/1 "%02x" " "' IR_1955.jpg | awk -f raw.txt > 16bitRAW.pgm
ore use imegamagick convert for saving a 16 Bit png

Code Select Expand

# hexdump -v -e '1/1 "%02x" " "' IR_1955.jpg | awk -f raw.txt | convert - IR_1955.png

finished...


The Script can also decode large raw datas over multiple Exif-Segments (max size is 65534Byte / segment)
There is a nice sample for testing from a very large 640 x 480 Flir Infrared Camera  P640 in wikepdia:
http://commons.wikimedia.org/wiki/File:Man_in_water_-_IR_image.jpg

in the picture the sensor raw datas  are shared over 10 JPEG-APP1-Blocks

Code Select Expand


> exiftool.exe -v3 Man_in_water_-_IR_image.jpg
... snip
JPEG APP1 (65532 bytes):
    1d7a: 46 4c 49 52 00 01 00 [b]09[/b] 46 46 46 00 00 00 00 00 [FLIR....FFF.....]  // 09 => 10 Blocks linked
    1d8a: 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 64 [...............d]
    1d9a: 00 00 00 40 00 00 00 0e 00 00 00 02 00 00 00 00 [...@............]
    1daa: 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 [................]
    1dba: 00 00 00 00 93 79 3c e2 00 01 00 02 00 00 00 65 [.....y<........e]
    1dca: 00 00 00 01 00 00 02 00 00 09 60 20 00 00 00 00 [..........` ....]
    1dda: 00 00 00 00 59 bc 40 05 00 20 00 01 00 00 00 6f [....Y.@.. .....o]
    [snip 65420 bytes]
  Warning = Ignored APP1 segment length 65532 (unknown header)
JPEG APP1 (65532 bytes):
   11d7a: 46 4c 49 52 00 01 01 09 78 3b 5e 3b 68 3b 50 3b [FLIR....x;^;h;P;]
   11d8a: 4d 3b 4d 3b 56 3b 61 3b 41 3b 39 3b 3a 3b 38 3b [M;M;V;a;A;9;:;8;]
   11d9a: 2a 3b 39 3b 2c 3b 06 3b 03 3b 15 3b 07 3b 0c 3b [*;9;,;.;.;.;.;.;]
   11daa: 0e 3b e5 3a 03 3b 1c 3b 4f 3b 2f 3b 08 3b ff 3a [.;.:.;.;O;/;.;.:]
   11dba: 11 3b 11 3b 29 3b 29 3b 31 3b 34 3b 2a 3b 53 3b [.;.;););1;4;*;S;]
   11dca: 53 3b 55 3b 5a 3b 60 3b 72 3b 6e 3b 5e 3b 76 3b [S;U;Z;`;r;n;^;v;]
   11dda: a8 3b bc 3b b3 3b 53 3b 30 3b 52 3b 58 3b 41 3b [.;.;.;S;0;R;X;A;]
    [snip 65420 bytes]
  Warning = Ignored APP1 segment length 65532 (unknown header)
JPEG APP1 (65532 bytes):
   21d7a: 46 4c 49 52 00 01 02 09 f4 3b ee 3b 05 3c 02 3c [FLIR.....;.;.<.<]
   21d8a: 02 3c 08 3c fe 3b 08 3c 09 3c 2f 3c 04 3c 16 3c [.<.<.;.<.</<.<.<]
   21d9a: f5 3b fe 3b f5 3b f4 3b 05 3c f1 3b fc 3b ec 3b [.;.;.;.;.<.;.;.;]
   21daa: ee 3b ee 3b f5 3b f5 3b ed 3b f2 3b f7 3b 05 3c [.;.;.;.;.;.;.;.<]
   21dba: e2 39 df 39 eb 39 ec 39 10 3a 57 3a 42 3a 33 3a [.9.9.9.9.:W:B:3:]
   21dca: 40 3a 5e 3a 78 3a 82 3a 81 3a a0 3a ac 3a 8b 3a [@:^:x:.:.:.:.:.:]
   21dda: b2 3a c3 3a 5d 3a 94 3a 9e 3a e8 3a bb 3a b6 3a [.:.:]:.:.:.:.:.:]
    [snip 65420 bytes]
  Warning = Ignored APP1 segment length 65532 (unknown header)
JPEG APP1 (65532 bytes):
   31d7a: 46 4c 49 52 00 01 03 09 a9 3a a6 3a 9b 3a 9e 3a [FLIR.....:.:.:.:]
   31d8a: bb 3a bb 3a bd 3a bc 3a b2 3a e2 3a dd 3a ba 3a [.:.:.:.:.:.:.:.:]
   31d9a: b4 3a b9 3a b5 3a b7 3a bf 3a ba 3a c1 3a ce 3a [.:.:.:.:.:.:.:.:]
   31daa: cb 3a d7 3a d0 3a c7 3a 74 3a 27 3a 3b 3a 15 3a [.:.:.:.:t:':;:.:]
   31dba: 2b 3a 23 3a 23 3a 12 3a 10 3a 39 3a 39 3a 63 3a [+:#:#:.:.:9:9:c:]
   31dca: a8 3a c0 3a be 3a 72 3a 08 3a e9 39 e7 39 01 3a [.:.:.:r:.:.9.9.:]
   31dda: 0c 3a 61 3a 2f 3a e6 39 a7 39 a3 39 a3 39 a2 39 [.:a:/:.9.9.9.9.9]
    [snip 65420 bytes]
  Warning = Ignored APP1 segment length 65532 (unknown header)
JPEG APP1 (65532 bytes):
   41d7a: 46 4c 49 52 00 01 04 09 4e 3b 5c 3b 61 3b 21 3b [FLIR....N;\;a;!;]
   41d8a: 09 3b 3a 3b d6 3a 64 3a 23 3b a4 3b ae 3b ac 3b [.;:;.:d:#;.;.;.;]
   41d9a: d3 3b c9 3b de 3b c4 3b b0 3b c6 3b c1 3b d2 3b [.;.;.;.;.;.;.;.;]
   41daa: ce 3b d4 3b cb 3b c4 3b cb 3b c8 3b 81 3b 7e 3b [.;.;.;.;.;.;.;~;]
   41dba: 85 3b a0 3b 7c 3b 74 3b 6d 3b 6b 3b 62 3b 69 3b [.;.;|;t;m;k;b;i;]
   41dca: 7f 3b 7f 3b 72 3b 5d 3b 58 3b 2d 3b 0b 3b 73 3b [.;.;r;];X;-;.;s;]
   41dda: 0e 3b 2a 3b 22 3b 2f 3b da 3a c1 3a cd 3a a8 3a [.;*;";/;.:.:.:.:]
    [snip 65420 bytes]
  Warning = Ignored APP1 segment length 65532 (unknown header)
JPEG APP1 (65532 bytes):
   51d7a: 46 4c 49 52 00 01 05 09 85 3a 84 3a 84 3a a1 3a [FLIR.....:.:.:.:]
   51d8a: a3 3a c1 3a da 3a cc 3a a6 3a e9 3a d6 3a a5 3a [.:.:.:.:.:.:.:.:]
   51d9a: 94 3a df 3a 08 3b 42 3b 97 3b ab 3b ac 3b 6c 3b [.:.:.;B;.;.;.;l;]
   51daa: 68 3b 80 3b 89 3b 79 3b 89 3b 78 3b 3c 3b 2b 3b [h;.;.;y;.;x;<;+;]
   51dba: 46 3b 41 3b 63 3b 91 3b 94 3b 96 3b 79 3b 77 3b [F;A;c;.;.;.;y;w;]
   51dca: 76 3b 99 3b 76 3b d6 3a 17 3b 48 3b 50 3b 5c 3b [v;.;v;.:.;H;P;\;]
   51dda: 6d 3b 6a 3b 6f 3b 78 3b 64 3b 93 3b 7c 3b 42 3b [m;j;o;x;d;.;|;B;]
    [snip 65420 bytes]
  Warning = Ignored APP1 segment length 65532 (unknown header)
JPEG APP1 (65532 bytes):
   61d7a: 46 4c 49 52 00 01 06 09 10 3b 06 3b f6 3a 09 3b [FLIR.....;.;.:.;]
   61d8a: 0b 3b 46 3b 79 3b cc 3b 9c 3b 8d 3b 6b 3b 94 3b [.;F;y;.;.;.;k;.;]
   61d9a: 85 3b 66 3b 5d 3b 6c 3b 93 3b 9a 3b aa 3b e4 3b [.;f;];l;.;.;.;.;]
   61daa: 92 3b b8 3b c8 3b b9 3b 8d 3b 70 3b 92 3b 9a 3b [.;.;.;.;.;p;.;.;]
   61dba: b2 3b 66 3b 5e 3b 69 3b 6c 3b 45 3b 45 3b 35 3b [.;f;^;i;l;E;E;5;]
   61dca: 2e 3b 67 3b dc 3b de 3b af 3b c6 3b cc 3b c9 3b [.;g;.;.;.;.;.;.;]
   61dda: 85 3b 32 3b 25 3b 87 3b b7 3b b0 3b 8c 3b 22 3b [.;2;%;.;.;.;.;";]
    [snip 65420 bytes]
  Warning = Ignored APP1 segment length 65532 (unknown header)
JPEG APP1 (65532 bytes):
   71d7a: 46 4c 49 52 00 01 07 09 98 3b 71 3b 5f 3b 47 3b [FLIR.....;q;_;G;]
   71d8a: 6d 3b c7 3b cb 3b cd 3b b0 3b a7 3b a1 3b c5 3b [m;.;.;.;.;.;.;.;]
   71d9a: 80 3b 87 3b 70 3b 6d 3b 92 3b ad 3b 86 3b c5 3b [.;.;p;m;.;.;.;.;]
   71daa: b4 3b af 3b c6 3b d6 3b d0 3b ac 3b b4 3b 8e 3b [.;.;.;.;.;.;.;.;]
   71dba: 91 3b bb 3b b1 3b b2 3b c4 3b c6 3b c9 3b a9 3b [.;.;.;.;.;.;.;.;]
   71dca: bb 3b ba 3b be 3b cb 3b b3 3b 9d 3b 9e 3b a6 3b [.;.;.;.;.;.;.;.;]
   71dda: a7 3b a3 3b a6 3b ae 3b 9b 3b 94 3b bb 3b bf 3b [.;.;.;.;.;.;.;.;]
    [snip 65420 bytes]
  Warning = Ignored APP1 segment length 65532 (unknown header)
JPEG APP1 (65532 bytes):
   81d7a: 46 4c 49 52 00 01 08 09 40 3b 5f 3b 45 3b 30 3b [FLIR....@;_;E;0;]
   81d8a: 27 3b 0c 3b 28 3b 51 3b 3d 3b 2a 3b 8e 3b 9f 3b [';.;(;Q;=;*;.;.;]
   81d9a: 51 3b 0e 3b 1d 3b 0c 3b 28 3b 35 3b 45 3b 06 3b [Q;.;.;.;(;5;E;.;]
   81daa: 07 3b 05 3b 08 3b 0f 3b f0 3a e8 3a f3 3a 06 3b [.;.;.;.;.:.:.:.;]
   81dba: 0f 3b 13 3b 13 3b fc 3a f5 3a f8 3a fc 3a d1 3a [.;.;.;.:.:.:.:.:]
   81dca: d0 3a c0 3a ef 3a d9 3a e2 3a 37 3b 4b 3b 4a 3b [.:.:.:.:.:7;K;J;]
   81dda: eb 3a ef 3a bb 3a bc 3a d6 3a e5 3a bc 3a 83 3a [.:.:.:.:.:.:.:.:]
    [snip 65420 bytes]
  Warning = Ignored APP1 segment length 65532 (unknown header)
JPEG APP1 (28644 bytes):
   91d7a: 46 4c 49 52 00 01 09 09 e4 3a fd 3a e8 3a f9 3a [FLIR.....:.:.:.:]
   91d8a: fe 3a 07 3b f8 3a fc 3a f4 3a fe 3a dc 3a e7 3a [.:.;.:.:.:.:.:.:]
   91d9a: d9 3a c3 3a c1 3a bd 3a ab 3a c5 3a c8 3a b2 3a [.:.:.:.:.:.:.:.:]
   91daa: d3 3a c7 3a a1 3a af 3a 94 3a 93 3a 92 3a 74 3a [.:.:.:.:.:.:.:t:]
   91dba: 7f 3a 86 3a ae 3a a4 3a b1 3a 97 3a b7 3a 9b 3a [.:.:.:.:.:.:.:.:]
   91dca: 9d 3a 91 3a 8a 3a 9e 3a 8f 3a 96 3a 93 3a a5 3a [.:.:.:.:.:.:.:.:]
   91dda: b3 3a b5 3a 96 3a b7 3a 7a 3a 96 3a 94 3a af 3a [.:.:.:.:z:.:.:.:]
    [snip 28532 bytes]
  Warning = Ignored APP1 segment length 28644 (unknown header)

as attachments the converted picture from wikipedia-site overlayed with 2 different palettes

Code Select Expand

# convert p640.png rain256.png -clut p640-rainbow.jpg
# convert p640.png Midgreen256.png -clut p640-Midgreen.jpg


[Phil Harvey]

Hi Thomas,

Thanks for this information.  I'll look at it in detail within the next few days and let you know if I have any questions.

- Phil

[tomas123]

another test pictures from wikipedia with full radiometric data in exif header

large 640x480 image from a Flir SC640
http://commons.wikimedia.org/wiki/File:IRWater.jpg

medium 320x240 image from a Flir E60
http://commons.wikimedia.org/wiki/File:IR_Fussbodenheizung.jpg

small 160x120 image from a Flir E30bx
http://commons.wikimedia.org/wiki/File:Image_thermique_de_l%27%C3%A9mission_d%27un_radiateur_%C3%A0_travers_un_mur.jpg


you can convert this samples with a one-liner by using your one colour palette
(use raw.txt from my first post)

Code Select Expand

$ hexdump -v -e '1/1 "%02x" " "' IR.jpg | awk -f raw.txt | convert - -auto-level rain256.png -clut converted.jpg
instead -auto-level (= -contrast-stretch 0%,0% ) you can use clipping of spikes for better results:  -contrast-stretch 2%,2%

as attachments some nice colour palettes for this one-liner ... and the three samples with rainbow palette

[Phil Harvey]

I haven't forgotten about this, but this will take a bit of work.  I'm hoping to get to it soon.

- Phil

[tomas123]

 no problem, it's a hint for you - my awk script works fine as workaround ... 

[Phil Harvey]

I have added the ability to extract the thermal image data.  It will appear in ExifTool 9.25.

Things went well.  There is lots of other information I would like to decode, but for now I am just extracting the thermal image data.

The only difference that may impact you is that I have formatted it as a TIFF image instead of PGM.

If you want to play with this before the official release, you can download a pre-release here.

- Phil

[tomas123]

wow, you knew that I'm using a mac osx binary...

all works fine

Code Select Expand

exiftool -b -FlirImage IR_2284.jpg | convert - -auto-level test.png
I tested it with a photo from a Flir E40 (120x160) and Flir SC640 (640x480).
Also successfull tested it with converted images from Flir Software "Flir Tools" and "Flir QuickReport".

Code Select Expand

exiftool -a -u -g1 IR_2262.jpg
...
---- FLIR ----
FLIR Image                      : (Binary data 38616 bytes, use -b option to extract)
FLIR Image Type                 : TIFF
better describe it with "FLIR 16 bit raw data"

[Phil Harvey]

Great!  Glad it works for you.

About the description... You're talking about the tag name?  (I can't change the "Binary data ..." description.)  I like FLIRImage because exiftool returns a valid image file format (either TIFF or PNG).  If I called it FLIRRawData then there would be no indication that the data is returned in a usable image format.

I am keen to try to decode more of the FLIR information, including the maker notes.  If you have FLIR hardware and/or software, maybe you can figure some of this out.  Use the -u option to see the MakerNotes information:

Code Select Expand

> exiftool -u -makernotes:all ../testpics/FLIR/Man_in_water_-_IR_image.jpg
Unknown 0x0001                  : 283
Unknown 0x0002                  : 276
Unknown 0x0003                  : 0.95
Unknown 0x0004                  : 238
Unknown 0x0005                  : 393
Unknown 0x0006                  : 233
Unknown 0x0007                  : 9C9E9EFA0A25446E21E3FFA54BCFD87
Unknown 0x0008                  : 00000000000000000000000000000000
Unknown 0x0009                  : 4........@oQ?Cp:.p:[...]
Unknown 0x000a                  : 1

Also, there are other records in the FLIR APP1 metadata that may have some interesting information.  Record number 0x01 contains the raw data ( I am identifying this by the record number instead of your technique of looking for 0x0200, but for the Man_in_water image for example there are also records 0x20, 0x22 and 0x21 (in that order) -- you can see these now with the -v3 option.

Thanks for any help you can provide in decoding more of this information.  I'll do what I can myself, but without FLIR equipment and software I probably won't be able to figure out very much.

- Phil

[tomas123]

I know, there are same magic temperatures fields in jpg.

as attachment a screenshot from (free) Flir Software with loaded "Man_in_water_-_IR_image.jpg"
Link: http://support.flir.com/SwDownload/app/RssSWDownload.aspx?ID=38

I'm only sure with this three tags:

Code Select Expand

Unknown 0x0003                  : 0.95 (adjust emissivity, see: http://en.wikipedia.org/wiki/Thermography#Emissivity )

Unknown 0x0005                  : 393 Kelvin (upper value of temperature range)
Unknown 0x0006                  : 233 Kelvin (lower value of temperature range)

the FLIR P640 has two temperature ranges:
  –40°C to +120°C (–40°F to +248°F)
  0°C to +500°C (+32°F to +932°F)

but if you are loading the Man_in_water in Flir Software above, then changes the values to

Code Select Expand

Unknown 0x0001                  : 300 K (max. Temp in Image)
Unknown 0x0002                  : 274K (min. Temp in Image)
Unknown 0x0003                  : 0.95 (emissivity)
Unknown 0x0004                  : 238
Unknown 0x0005                  : 292K ??
Unknown 0x0006                  : 275K ??


I think its not worth to decode this values, but you can export the decoded sensor size (16 bit tiff)

[Phil Harvey]

This is very useful, thanks!  I didn't know about the free software.  I'll download it and run some tests when I can get access to a Windows system (it will be a few days).

- Phil

[Phil Harvey]

I like FLIRImage because exiftool returns a valid image file format (either TIFF or PNG).  If I called it FLIRRawData then there would be no indication that the data is returned in a usable image format.

... or maybe RawThermalImage?

- Phil

[tomas123]

wow, you knew that I'm using a mac osx binary...

your binary is a perl script   

I looked at your FLIR.pm and saw, that you also decoded the strange PNG files from Flir B50, B60, P60, i7 Etc. 

RawThermalImage is nice

some background infos:
the dependency between 16 Bit RAW and temperature is not 100% linear.
the Flir software calculates between sensor values and some physical environment conditions
google:    flir "The measurement formula"

on a sample picture I got following correlation between RAW and Grad Celsius.
my temperature range is 0 = –40°C and 2^16  = +120°C

Code Select Expand


RAW    Grad Celsius
0 -40,011
18498 25,791
18598 26,319
18722 26,969
18861 27,694
18980 28,311
19080 28,826
19180 29,34
19280 29,851
19380 30,359
19480 30,865
19580 31,369
19680 31,871
19782 32,381
19884 32,888
19992 33,423
20104 33,975
20210 34,495
20315 35,008
20433 35,581
20567 36,23
20721 36,97
20928 37,958
21121 38,872
21379 40,083
21620 41,204
21813 42,094
21985 42,882
22114 43,47
22270 44,177
22417 44,839
22611 45,709
22891 46,953
23446 49,385
24636 54,454

the batch decoding of raw values is very usefull for making of panorama pictures (no scale and logo on the single picture)
the resolution of a "low cost" infrared cam is awfull
but I you make 4x4 pictures (or more), then you get a nice upgrade...

[tomas123]

a comment:

Code Select Expand

ExifTool.pm
4932                my $chunkNum = Get8u($segDataPt, 6);
4933                my $chunksTot = Get8u($segDataPt, 7) + 1; # (note the "+ 1"!)
4934                $verbose and print $out "$$self{INDENT}FLIR chunk $chunkNum of $chunksTot\n";

this gives from
  FLIR chunk 0 of 10
to last
  FLIR chunk 9 of 10

[tomas123]

it's crazy:
for decoding the PNG you must swap upper and lower byte (little endian)

link to cat picture:
http://www.nuage.ch/site/flir-i7-some-analysis/

Code Select Expand

$ exiftool -b -FlirImage IR_0248.jpg | convert - -compress none cat.pgm
$ awk -f png.txt cat.pgm | convert -  -auto-level cat.png
$ cat png.txt
{
if (NR < 4)
   { print $0 }
else
   { for(i=1; i <= NF ; i++)
     {
      k = 256*($(i)%256)+int($(i)/256)
      printf("%d ",k)
     }
     print ""
   } 
}

see attachment for result

a larger PNG (180x180) from Flir B60
http://commons.wikimedia.org/wiki/File:Aqua_Tower_thermal_image.jpg

[Phil Harvey]

your binary is a perl script   

Yes.  All I needed to know was that you weren't running the Windows EXE version.  The other versions are pure Perl.

I looked at your FLIR.pm and saw, that you also decoded the strange PNG files from Flir B50, B60, P60, i7 Etc. 

Yes.  I noticed the funny colours too, but I understand now that you tell me they swapped the byte order.  Funny.

RawThermalImage is nice

Great.  I'll go with that then.

some background infos:
the dependency between 16 Bit RAW and temperature is not 100% linear.
the Flir software calculates between sensor values and some physical environment conditions
google:    flir "The measurement formula"

Thanks for the info.

this gives from
  FLIR chunk 0 of 10
to last
  FLIR chunk 9 of 10

Yes.  I wasn't entirely happy about this either, but it is accurate.  The chunks are numbered from 0.  However, it does make sense for me to add 1 for the print statement.  I'll think about this.

- Phil