video codec as voiceband modem bitrates

[tomasparks]

I've been test a group of video codec at voice band modem bitrates
tried h.261, h.263,mpeg 1,mpeg 2,mpeg-4 part 2, vp9, mjpeg

H.264 test:

300b,1.2k,9.6kb forget it I cant get the ffmpeg to work at such a low bitrate
14.4kb Unwatchable
28.8kb watchable lots of compression artifact
33.6kb watchable lots of compression artifact
50kb watchable acceptable compression artifacts
100kb (two 56kb bound) watchable nearly visually lossless except when there is camera pan and/or fast motion

I get 1/2kb above/bellow requested bitrate

VP9 look good at 50kbits

VP9:
ffmpeg -i "video_input" -deinterlace -r 12.5 -s 320x240 -crf 0 -b:v 50000 -minrate 50000 -maxrate 50000 -bufsize 50k -threads 8 -auto-alt-ref 1 -lag-in-frames 12 -tile-columns 6 -frame-parallel 1 -speed 0 -an -sn "video_output-50k.webm"

H.264
ffmpeg -i "video_input" -deinterlace -profile:v baseline -level 3.0 -pix_fmt yuv420p -movflags +faststart -r 12.5 -s 320x240 -c:v libx264 -crf 31 -v:b 50000 -maxrate 50000 -bufsize 50k -an -sn -preset veryslow "video_output-50k.mkv"

source video was from DVB-T (mpeg-2)

[smok3]

Min resolution to be consider is 320x240 and 12,5 fps?

[pandy]

What do you expect with this resolution (CIF - https://en.wikipedia.org/wiki/Common_Intermediate_Format ) and such high framerate? - most teleconference codecs assume at least 1 ISDN line (so at least 64kbps).

[tomasparks]

Daala codec not plug and play :(
Dirac codec (Schrödinger) 4 years old seams dead

I tried H.261 to get the codec to work at at 50kb, I had to use QCIF (176x144) :(

i've been testing vp9 with 320x240 16:9 1fps

i've tried:
* 9.6kb the resulting bitrate is about 10.4kb, looks just ok
* 14.4k, look acceptable at 640x480 (doubled frame size)

[benwaggoner]

You should also include HEVC, which does well at those low bitrates.

[filler56789]

So you all dared to forget RealVideo? :)

[pandy]

So you all dared to forget RealVideo? :)

Nope, but RV was VFR - not sure about H.264 capabilities on VFR - telecom video!= broadcast video - with aggressive temporal preprocessing (ffmpeg https://www.ffmpeg.org/ffmpeg-filters.html#thumbnail or stacked https://www.ffmpeg.org/ffmpeg-filters.html#select_002c-aselect with scene detection) it should be not a problem to match and perhaps outperform RealVideo

[filler56789]

Nope, but RV was VFR - not sure about H.264 capabilities on VFR - telecom video!= broadcast video

:goodpost: I stand corrected, sir.

[tomasparks]

sorry for the long delay in replying:

I've been doing a bit research on this project and i've come to the conclusion.

I am looing for Progressive/Incremental Transmission Compression formats, witch NASA is the only one using a work version on Mars :D but it is for still images only :(


The ICER Progressive Wavelet Image Compressor
A. Kiely 1 and M. Klimesh 1
(NASA)
ICER is a progressive, wavelet-based image data compressor designed to meet
the specialized needs of deep-space applications while achieving state-of-the-art
compression effectiveness. ICER can provide lossless and lossy compression, and
incorporates an error-containment scheme to limit the effects of data loss during
transmission. The Mars Exploration Rover (MER) mission will rely primarily on
a software implementation of ICER for image compression. This article describes
ICER and the methods it uses to meet its goals, and explains the rationale behind
the choice of methods. Performance results also are presented.


Image Compression for Progressive Transmission
V.G. Ruiz, J.J. Fernández, I. Garcı́a
In this work several different strategies for image compression and progressive transmission are
analyzed and evaluated. It will be shown that the simplest wavelet transform (Unitary Sequential
Transform, UST) with a spatial predictor (U(S+P)T) followed by the so-called Set Partitioning In
Hierarchical Trees (SPIHT) coder is a very good lossless compressor providing compression rates
similar to CALIC or LOCO-I. In addition, U(S+P)T+SPIHT is an optimal solution for image
progressive transmission because the most relevant information is concentrated in a small amount
of data.

[pandy]

Don't forget about https://en.wikipedia.org/wiki/Slow-scan_television - it is even more bandwidth friendly than regular modem... :cool:

[benwaggoner]

sorry for the long delay in replying:

I've been doing a bit research on this project and i've come to the conclusion.

I am looing for Progressive/Incremental Transmission Compression formats, witch NASA is the only one using a work version on Mars :D but it is for still images only :(

Anything that's working on Mars is probably somewhat out of date :).

Scalable HEVC?

https://hevc.hhi.fraunhofer.de/shvc