3

In attempting to record in the highest quality on a JVC GC-PX100 camcorder, I've been recording 1080p video. Unfortunately, viewing the video upon edit shows very strong compression artifacts, which affect quality to the point that burning to DVD and burning to Blu-Ray end up with roughly the same results. The video was taken on a reasonable photo tripod, so the camera was definitely steady, and the bulk of the video was unmoving walls with detailed paintings.

Adobe Premiere Elements 11 is what I'm doing for editing, and I've just tried the Neat Video plugin. Unfortunately, while the default settings reduce the artifacts/noise in flat areas (cheeks, walls), but not along all edges (faces).

Essentially, the video is "crunchy", and that's visible on playback on a large TV.

Adobe Elements shows the .mts file from the camcorder is MPEG Movie, 1920x1080, frame rate 59.94, 48kHz audio, average data rate 3.3 MB/s (~34Mbit/s), pixel aspect ration 1.0.

VLC shows the .mts file from the camcorder is Codec H264 - MPEG-R AVC (part 10)(h264), 1920x1080, frame rate 59.940060, decoded format Planar 4:2:0 YUV, with A52 (aka AC3) 48kHz audio.

My goal is the highest quality result video, targeting Blu-Ray output. My budget can handle a plugin fee, but not a new camcorder (spent too much on the add-on microphone).

So my questions in particular are:

  • Are there better settings to increase overall quality during recording?
  • Are there better settings for Neat Video
  • Is there another way to reduce compression artifacts in HD video taken by a good consumer camcorder?

EDITED TO ADD: AJ, Jim, thank you for your answers; they led me to the regrettable conclusion that my expectations are too high for any 1080p60 mode at 35Mbit, and no 1080p30/25/24 mode is available on my model.

Even after the usual firmware update, and a factory reset as recommended by JVC Customer Support, the video is "crunchier" than I like. There is no 1080p30 mode, unfortunately, and lesser modes have lesser bitrates as well.

After some experimentation, it appears that the AVCHD and MP4 1080p60 modes are more or less equivalent, with a possible slight edge to the AVCHD variant.

NeatVideo needs to have the plugin expanded and the Temporal Radius increased, as well as going into the settings to profile the video clip in question. Edges remain crunchy, and dedicated "reference" profile video appears necessary to get good results.

3
  • You could try to make your video more compressible by using a tripod, or trying to hold the camera more steady, if that's an option. Then the bad hardware encoder should have an easier time finding similarity with previous frames, and not having to throw away as much information. Commented Jan 15, 2015 at 14:27
  • It was tripod mounted, but thank you; I'll update the question. Commented Jan 16, 2015 at 5:18
  • Wow, that's just horrible if it has bad artifacting on low-motion content. Is it recording in intra-only mode? ffplay -vf showinfo, and look at the frame types. If you see any type:P or type:B, then most of the blocks in each frame should just reference previous frames. Sounds like a bad choice of chips by JVC, assuming there are some HW encoders that don't suck quite as much, compared to x264. Commented Jan 16, 2015 at 5:34

2 Answers 2

2

There is no way around needing more data rate to capture that much video information. It sounds like the camera probably isn't actually capable of delivering 1080P/60 but that they use a very low quality version so they can slap the label on the box.

Data rates are generally measured in Mb(bits)/sec rather than MB/s, but even if your suggested rate actually is in MB/sec, 3.3MB/s for the video you are describing is near uselessly small and strong artifacting would be expected (particularly for a real time encoding) as the compression algorithm breaks down.

You need to use a mode that supports either a higher data rate or supplies less data to be encoded at that data rate. There is no other real solution since there is no information there to actually use to fix the problem.

1
  • Thank you. I'll try the 720p modes as well, since higher resolution is counterproductive if compression destroys more detail than is gained, though from the spec sheet they record at a lower bit rate (MOV: 1920x1080/60p, 40Mbps (Video: 36Mbps, Audio: 4Mbps) MP4: 1920x1080/60p, 36Mbps / 1280x720/60p, 16Mbps iFrame: 1280x720/30p, 36Mbps AVCHD Progressive: 1920x1080/60p, 28Mbps AVCHD: 1920x1080/60i, 17Mbps/5Mbps MP4 (High Speed) 120/240/300fps: 640x360, 420/600fps: 320x176). Commented Mar 31, 2014 at 15:17
3

I assume the actual data rate is 3.3 Mb/s, not MB/s (bits, not bytes). This is very low for 1080p60, and artifacts are likely at such a low rate. You might consider capturing at at least double that rate.

Alternatively (or also), halving the frame rate to 29.97 would double the effective bit rate. You would be trading some motion artifacting for less visual artifacting, but the former is usually much less noticeable.

Considering that blu-ray can handle in the range of 40-50 Mb/s, your bit rate is way below the target for the delivery medium you've chosen.

4
  • Even if it was 3.3 MB/s, it is still very low for 1080p at 60fps. Uncompressed that'd be 124 MB a second, so 3 is a 40 to 1 compression ratio...
    – AJ Henderson
    Commented Mar 31, 2014 at 14:15
  • I've updated the question: The data rate really is in Megabytes Per Second, and 3.3MB/s matches fairly well with JVC's claims of 35 Megabits per second. I'll try and see if recording in 1080p30 still records at the same data rate, and if it reduces artifacts. Commented Mar 31, 2014 at 15:03
  • Ah, well 3.3 MB/s is more like 26Mb/s (3.3 x 8), which I would expect to provide fair video quality. I do think you'll get better pictures at 30p, but only testing in your environment will tell. You may also find that the sensor isn't really providing 1920x1080 and that 720p looks better overall.
    – Jim Mack
    Commented Mar 31, 2014 at 17:40
  • hardware h.264 encoders are usually horrible compared to even x264 -preset superfast. Commented Jan 15, 2015 at 14:24

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge you have read our privacy policy.

Not the answer you're looking for? Browse other questions tagged or ask your own question.