How to extract 16-bit PNG frame from lossless x264 video

Question

I encoded a series of 16-bit grayscale PNGs to a lossless video with the following command:

ffmpeg -i image%04d.png -crf 0 -c:v libx264 -preset veryslow output.mp4

I am now trying to verify that the conversion to video was truly lossless by pulling out the PNGs at the same quality. The command I'm using:

ffmpeg -i output.mp4 image%04d.png

However, this is outputting 8-bit PNGs. I've tried various options I've read about such as -vcodec png and -qscale 0 but so far nothing appears to make it output 16-bit PNGs.

How do I extract all frames from the video at the same quality as they were going in? Or did I make a mistake in creating the lossless video in the first place?

Edit: I get this error message when trying to use -pix_fmt gray16be.

[swscaler @ 0x7fef1a8f0800] deprecated pixel format used, make sure you did set range correctly

Full output:

ffmpeg -i output.mp4 -pix_fmt gray16be  image%04d.png
ffmpeg version 3.3.1 Copyright (c) 2000-2017 the FFmpeg developers
  built with Apple LLVM version 8.0.0 (clang-800.0.42.1)
  configuration: --prefix=/usr/local/Cellar/ffmpeg/3.3.1 --enable-shared --enable-pthreads --enable-gpl --enable-version3 --enable-hardcoded-tables --enable-avresample --cc=clang --host-cflags= --host-ldflags= --enable-libmp3lame --enable-libx264 --enable-libxvid --enable-opencl --disable-lzma --enable-vda
  libavutil      55. 58.100 / 55. 58.100
  libavcodec     57. 89.100 / 57. 89.100
  libavformat    57. 71.100 / 57. 71.100
  libavdevice    57.  6.100 / 57.  6.100
  libavfilter     6. 82.100 /  6. 82.100
  libavresample   3.  5.  0 /  3.  5.  0
  libswscale      4.  6.100 /  4.  6.100
  libswresample   2.  7.100 /  2.  7.100
  libpostproc    54.  5.100 / 54.  5.100
Input #0, mov,mp4,m4a,3gp,3g2,mj2, from 'output.mp4':
  Metadata:
    major_brand     : isom
    minor_version   : 512
    compatible_brands: isomiso2avc1mp41
    encoder         : Lavf57.71.100
  Duration: 00:00:09.76, start: 0.000000, bitrate: 1337 kb/s
    Stream #0:0(und): Video: h264 (High 4:4:4 Predictive) (avc1 / 0x31637661), yuvj444p(pc), 512x512, 1336 kb/s, 25 fps, 25 tbr, 12800 tbn, 50 tbc (default)
    Metadata:
      handler_name    : VideoHandler
Stream mapping:
  Stream #0:0 -> #0:0 (h264 (native) -> png (native))
Press [q] to stop, [?] for help
[swscaler @ 0x7fef1a8f0800] deprecated pixel format used, make sure you did set range correctly
Output #0, image2, to 'image%04d.png':
  Metadata:
    major_brand     : isom
    minor_version   : 512
    compatible_brands: isomiso2avc1mp41
    encoder         : Lavf57.71.100
    Stream #0:0(und): Video: png, gray16be, 512x512, q=2-31, 200 kb/s, 25 fps, 25 tbn, 25 tbc (default)
    Metadata:
      handler_name    : VideoHandler
      encoder         : Lavc57.89.100 png
frame=  244 fps=0.0 q=-0.0 Lsize=N/A time=00:00:09.76 bitrate=N/A speed=  21x    
video:4038kB audio:0kB subtitle:0kB other streams:0kB global headers:0kB muxing overhead: unknown

I'm happy to use a non-ffmpeg solution if there is one.

Answer 1

Your lossless video is 8-bit, as that's what x264 has encoded to. And in any case, x264 does not go above 10-bit (for which you would need a standalone encoder, or a different ffmpeg binary).

-c:v ffv1 is lossless and can encode to 16-bit gray (gray16le) or 16-bit RGB (e.g. rgb48le)

---

That swscaler line is a warning, not an error. FFmpeg does report the output as 16-bit per channel,

Stream #0:0(und): Video: png, gray16be, 512x512 ...

but you've already lost fidelity with the downgrade of the bit depth when you encoded to H.264.

I'm unclear on what the pixel format of the input PNGs is - RGB with no hint of hue, or a monochannel encoding? I suspect the former, in which case, use

ffmpeg -i image%04d.png -c:v ffv1 output.nut

Then you can do a compare with

ffmpeg -i output.nut -i image%04d.png -filter_complex ssim -f null -

A readout of

SSIM R:1.000000 (inf) G:1.000000 (inf) B:1.000000 (inf) All:1.000000 (inf)

indicates full fidelity.