Multiple format changes in FFmpeg for thermal camera

I'm having trouble generating a command to process output from a uvc thermal camera's raw data so that it can be colorized and then output to a virtual device with the intention of streaming it over rtsp. This is on a raspberry pi 3B+ with 32bit bullseye.

The original code that works perfectly for previewing it is:

ffmpeg -input_format yuyv422 -video_size 256x384 -i /dev/video0 -vf 'crop=h=(ih/2):y=(ih/2)' -pix_fmt yuyv422 -f rawvideo - | ffplay -pixel_format gray16le -video_size 256x192 -f rawvideo -i - -vf 'normalize=smoothing=10, format=pix_fmts=rgb48, pseudocolor=p=inferno'

Essentially what this is doing is taking the raw data, cutting the useful portion out, then piping it to ffplay where it is seen as 16bit grayscale (in this case gray16le), then it is normalized, formatted to 48 bit rgb and then a pseudocolor filter is applied.

I haven't been able to get this to translate into ffmpeg-only because it throws codec errors or format errors or converts the 16bit to 10bit even though I need the 16bit. I have tried using v4l2loopback and two instances of ffmpeg in separate windows to see if I could figure out where the error was actually occuring but I suspect that is introducing more format issues that are distracting from the original problem. The closest I have been able to get is

ffmpeg -input_format yuyv422 -video_size 256x384 -i /dev/video0 -vf 'crop=h=(ih/2):y=(ih/2)' -pix_fmt yuyv422 -f rawvideo /dev/video3


Followed by

ffmpeg -video_size 256x192 -i /dev/video3 -f rawvideo -pix_fmt gray16le -vf 'normalize=smoothing=10,format=pix_fmts=rgb48, pseudocolor=p=inferno' -f rawvideo -f v4l2 /dev/video4

This results in a non colorized but somewhat useful image with certain temperatures showing as missing pixels as opposed to the command with ffplay where it shows a properly colorized stream without missing pixels.

I'll include my configuration and log from the preview command but the log doesn't show errors unless I try to modify parameters and presumably mess up the syntax.

 ffmpeg -input_format yuyv422 -video_size 256x384 -i /dev/video0 -vf 'crop=h=(ih/2):y=(ih/2)' -pix_fmt yuyv422 -f rawvideo - | ffplay -pixel_format gray16le -video_size 256x192 -f rawvideo -i - -vf 'normalize=smoothing=10, format=pix_fmts=rgb48, pseudocolor=p=inferno'
ffplay version N-109758-gbdc76f467f Copyright (c) 2003-2023 the FFmpeg developers
  built with gcc 10 (Raspbian 10.2.1-6+rpi1)
  configuration: --prefix=/usr/local --enable-nonfree --enable-gpl --enable-hardcoded-tables --disable-ffprobe --disable-ffplay --enable-libx264 --enable-libx265 --enable-sdl --enable-sdl2 --enable-ffplay
  libavutil      57. 44.100 / 57. 44.100
  libavcodec     59. 63.100 / 59. 63.100
  libavformat    59. 38.100 / 59. 38.100
  libavdevice    59.  8.101 / 59.  8.101
  libavfilter     8. 56.100 /  8. 56.100
  libswscale      6.  8.112 /  6.  8.112
  libswresample   4.  9.100 /  4.  9.100
  libpostproc    56.  7.100 / 56.  7.100
ffmpeg version N-109758-gbdc76f467f Copyright (c) 2000-2023 the FFmpeg developers
  built with gcc 10 (Raspbian 10.2.1-6+rpi1)
  configuration: --prefix=/usr/local --enable-nonfree --enable-gpl --enable-hardcoded-tables --disable-ffprobe --disable-ffplay --enable-libx264 --enable-libx265 --enable-sdl --enable-sdl2 --enable-ffplay
  libavutil      57. 44.100 / 57. 44.100
  libavcodec     59. 63.100 / 59. 63.100
  libavformat    59. 38.100 / 59. 38.100
  libavdevice    59.  8.101 / 59.  8.101
  libavfilter     8. 56.100 /  8. 56.100
  libswscale      6.  8.112 /  6.  8.112
  libswresample   4.  9.100 /  4.  9.100
  libpostproc    56.  7.100 / 56.  7.100
Input #0, video4linux2,v4l2, from '/dev/video0':B sq=    0B f=0/0   
  Duration: N/A, start: 242.040935, bitrate: 39321 kb/s
  Stream #0:0: Video: rawvideo (YUY2 / 0x32595559), yuyv422, 256x384, 39321 kb/s, 25 fps, 25 tbr, 1000k tbn
Stream mapping:.000 fd=   0 aq=    0KB vq=    0KB sq=    0B f=0/0   
  Stream #0:0 -> #0:0 (rawvideo (native) -> rawvideo (native))
Press [q] to stop, [?] for help
Output #0, rawvideo, to 'pipe:':   0KB vq=    0KB sq=    0B f=0/0   
  Metadata:
    encoder         : Lavf59.38.100
  Stream #0:0: Video: rawvideo (YUY2 / 0x32595559), yuyv422(tv, progressive), 256x192, q=2-31, 19660 kb/s, 25 fps, 25 tbn
    Metadata:
      encoder         : Lavc59.63.100 rawvideo
frame=    0 fps=0.0 q=0.0 size=       0kB time=-577014:32:22.77 bitrate=  -0.0kbInput #0, rawvideo, from 'fd:':    0KB vq=    0KB sq=    0B f=0/0   
  Duration: N/A, start: 0.000000, bitrate: 19660 kb/s
  Stream #0:0: Video: rawvideo (Y1[0][16] / 0x10003159), gray16le, 256x192, 19660 kb/s, 25 tbr, 25 tbn
frame=   13 fps=0.0 q=-0.0 size=    1152kB time=00:00:00.52 bitrate=18148.4kbitsframe=   25 fps= 24 q=-0.0 size=    2304kB time=00:00:01.00 bitrate=18874.4kbitsframe=   39 fps= 25 q=-0.0 size=    3648kB time=00:00:01.56 bitrate=19156.7kbitsframe=   51 fps= 24 q=-0.0 size=    4800kB time=00:00:02.04 bitrate=19275.3kbitsframe=   64 fps= 24 q=-0.0 size=    6048kB time=00:00:02.56 bitrate=19353.6kbitsframe=   78 fps= 25 q=-0.0 size=    7392kB time=00:00:03.12 bitrate=19408.7kbits

I'd also like to use the correct option so it isn't scrolling though every frame in the log as well as links to resources for adapting a command to a script for beginners even though that's outside the purview of this question so any direction on those would be much appreciated.

@gyan 's question gave me an idea and I realized that the original code was given in stages to show the theory behind what was happening and I was over complicating it. I was able to get it working with

ffmpeg -input_format gray16le -video_size 256x384 -i /dev/video0 -vf 'crop=h=(ih/2):y=(ih/2), normalize=smoothing=10, pseudocolor=p=inferno' -pix_fmt yuv420p -f v4l2 /dev/video2