-I have an Xvid with a resolution of [576 x 432] at about [1470 kb/s]. In order to test what Mpeg2 bitrate would be overkill, I've been doing CQ encodes with TMPGEnc. However, determining this point has been difficult because even at CQ 100 with a Min of [1000] and a Max of [9800], I still get some blockiness when the picture stops. I opened up VirtualDubMod and opened the original Xvid [.avi] and opened the produced [.m2v]. The Mpeg4 has no such blockiness on any of these frames, but when there's motion, I cannot tell the difference as much.
-Is it not supposed to be the other way around? Shouldn't I see more imperfections at higher motion scenes than at lower motion scenes? The final average bitrate came out to 7100 (something), and the Min never dropped below 4000 - I checked this with the [.log] file.
-From what I've heard from this site and others, CQ 100 is more than overkill. What then is the cause of the problem? Blockiness is only supposed to be a problem at lower (VCD) bitrates.
I plugged in these settings to eliminate any variables; so I could compare the bitrates of two identical videos with the exception of the fact that one was done with Xvid compression while the other was done with Mpeg2 compression. Here are my CQ Test settings:
-Mpeg2
-576x432
-1:1 VGA
-23.976 fps
-Non-interlace
-9 Bit DC Comp. Prec.
-Highest Quality Motion Search
-GOP = 1-5-2
-18 = Max # of Frames in GOP
-Detect Scene Change
-CG/Animation Quantization Matrix (Source is a cartoon - japanese animation)
-Use Floating Point DCT
+ Reply to Thread
Results 1 to 8 of 8
-
-
It seems to do this even when I encode down from [576 x 432] to [512 x 384]. For testing purposes, I am now doing a CBR of 9800. I still get blockiness in the image, blockiness I don't see in the original image.
-
I opened it up with TMPGEnc direct rather than frameserving through VirtualDubMod and got much less artifacts. Funny, though. I always thought that VirtualDub was better at decompressing the video than TMPGEnc...
Oh, and I guess I just worded it wrong. I meant to say when I looked at frame by frame. -
I also tried AviSynth, which I believe uses Huffyuv to decompress the video. It still left a few more artifacts than opening in TMPGEnc direct. When TMPGEnc encodes, does it not decompress frame by frame as it encodes like a frameserver decompresses frame by frame as it delivers the video to the encoder?
At any rate, it seems that decompression of the frameservers and the fact that I had the Quantization Matrix on CG/Animation rather than leaving it at Default is what led me to have these artifacts. -
They are to me. You can see it easier when you pause it or see it frame by frame, but you can still notice it when the picture plays at full speed. I even did a 2-Pass VBR of 9800-9800-9800 (min, avg, max). I still saw the same artifacts I was seeing at lower bitrates. I was trying to determine actual overkill by using CQ with TMPGEnc, but I wouldn't be able to see that if I see artifacts at such a high bitrate. Like I said, when I open it with TMPGEnc direct, I notice significantly lower artifacts than when I frameserve with AviSynth or VirtualDub, but some artifacts are going to exist. Perhaps, this is just decompression imperfection.
-
Ok, I think VirtualDub isn't causing any artifact problems. It looks like it looked that way when I viewed it in VirtualDub using Preview Input button, but not otherwise.
Similar Threads
-
Converting YV12 to anything else causes inexplicable blockiness
By Asterra in forum Capturing and VCRReplies: 13Last Post: 19th Apr 2011, 06:02 -
m2ts to xvid or h264 always gets blockiness
By ubulord in forum Video ConversionReplies: 29Last Post: 24th Dec 2010, 12:18 -
Simulate Video Streaming Problems (Tearing, Blockiness, Smearing, etc.)
By TheMcD in forum EditingReplies: 14Last Post: 21st Oct 2010, 05:55 -
Blockiness in a running water scene
By Umen Pich in forum Video ConversionReplies: 10Last Post: 20th Nov 2008, 10:32 -
problem encoding error decompressing video frame 0 :
By goingape in forum Newbie / General discussionsReplies: 7Last Post: 27th May 2007, 13:14