On Thu, 22 Jul 2004, Marc Gregoire wrote:
> I took another image of which I do have the original to work with.
> This picture can be found at the following links:
> Original: http://www.nuonsoft.com/temp/07_orig.jpg
> Scaled: http://www.nuonsoft.com/temp/07_scaled.jpg
> Mpeg2enc result: http://www.nuonsoft.com/temp/07_mpeg2enc.png
> Commercial result: http://www.nuonsoft.com/temp/07_commercial.png
Thanks for making the data available - I pulled the images in and
did a bit of experimenting tonight. The results were even better
than I expected!
> 1360x1020. The camera automatically saves them to a low (to avoid
> graphical artifacts) compressed JPG.
> You clearly see that the 07_mpeg2enc.png shot is more blocky than the
> 07_commercial.png. This is especially noticeable in the lower
> left corner of the image. You can see block in the marble. You also see
> blocks in the orange wall in the back. These blocks are not visible in the
To my eyes the blocks are moderately hard to see - they're there
but on a TV I'm certain noone would be able to tell the two apart
easily.
> took the jpeg2yuv of the mjpegtools 1.6.2 and that version does indeed have
> the -R option, but like you said, this didn't fix the quality issue.
You can also run the data thru "yuvcorrect -Y CONFORM" and it will
adjust the data to be within the broadcast safe range.
Now on to the evening's testing. What I was curious about was if
doing the scaling and color space conversion with 'y4mscaler' would
make a difference (hopefully for the better ;)).
So, I took the 07_orig.jpg file and converted it to 24bit RGB PPM
and used 07_orig.ppm as the source. 'ppmtoy4m' was used to convert
to 4:4:4 Y4M format (lossless - just going from RGB to YUV, still in
24bit mode).
Then the key step: I used y4mscaler with an 8 tap Sinc/Lanczos scaling
kernel _and_ converted to 420_mpeg2 *and* took into account the fact
that PAL pixels are really 59:54 and not 1:1 (if you just scale from
1360x1020 to 720x576 you're changing the aspect ratio when displayed
on a TV!).
The last step was to run the data thru the yuvcorrect program to get
broadcast safe values. The result of this could have been piped
directly into mpeg2enc but I created a .y4m file that I could use with
other programs.
Encoding was done with "-q 1" and a bitrate of 9800. The resulting
.m2v file was ~84K bytes ("84095 22 Jul 19:52 xxx.m2v")
To view the results it's necessary to decode the .m2v file back to PPM.
That is done using 'mpeg2dec' and a couple more of the mjpegtools
utilities (pgmtoy4m and y4mtoppm). y4mscaler was used to convert
upsample to 4:4:4 (24bit mode). pgmtoy4m is I believe only in the
CVS version of mjpegtools at the moment but I might be mistaken about
that (it's been eons since I've used anything except the CVS version ;))
The script I used was this:
----------
#!/bin/sh
cat 07_orig.ppm | \
ppmtoy4m -n 1 -F 25:1 -I p -S 444 | \
y4mscaler -I sar=1:1 -O sar=PAL -O size=720x576 -O chromass=420_mpeg2 -S
option=sinc:8 | \
yuvcorrect -Y CONFORM > 07_orig.y4m
mpeg2enc -f 8 -b 9800 -q 1 -o xxx.m2v < 07_orig.y4m
mpeg2dec -o pgmpipe xxx.m2v | \
pgmtoy4m -a 59:54 -r 25:1 | \
y4mscaler -O chromass=444 | \
y4mtoppm -L > 0.ppm
----------
I have attached a script of the session so that you can see the
output of the tools. The output from y4mscaler will be of interest I
believe.
Maybe I'm overdue for getting new glasses but to me the .m2v file
produced above looks sharper and clearer than either 07_commercial.png
or 07_mpeg2enc.png. The differences between the file produced above
and the .png files are subtle to my eyes even at a 2x zoom in the
GIMP (using Apple's 23" Cinema-HD LCD screen). I'm sure that With a
higher bitrate the differences would be even smaller but the
sharpness in the file created by the above script is noticeably
better than the 07_commercial.png and 07_mpeg2enc.png - those almost
look "blurred" in comparison ;)
Perhaps someone with better eyes (or newer glasses :)) can run some
tests and chime in on this - another opinion would be welcome.
Cheers,
Steven Schultz
Script started on Thu Jul 22 19:52:03 2004
[LARRY:~/x] sms% ./x.sh
INFO: [yuvcorrect] yuvcorrect 1.6.3 (16-11-2002) is a general image correction
utility for yuv frames
INFO: [yuvcorrect] (C) 2002 Xavier Biquard <[EMAIL PROTECTED]>, yuvcorrect -h for
usage, or man yuvcorrect
INFO: [ppmtoy4m] Command-line Parameters:
INFO: [ppmtoy4m] framerate: 25:1
INFO: [ppmtoy4m] pixel aspect ratio: 1:1
INFO: [ppmtoy4m] interlace: none/progressive
INFO: [ppmtoy4m] starting frame: 0
INFO: [ppmtoy4m] # of frames: 1, or until input exhausted
INFO: [ppmtoy4m] chroma subsampling: 4:4:4 (no subsampling)
INFO: [ppmtoy4m] Output Stream parameters:
INFO: [ppmtoy4m] frame size: 1360x1020 pixels (4161600 bytes)
INFO: [ppmtoy4m] chroma: 4:4:4 (no subsampling)
INFO: [ppmtoy4m] frame rate: 25/1 fps (~25.000000)
INFO: [ppmtoy4m] interlace: none/progressive
INFO: [ppmtoy4m] sample aspect ratio: 1:1
INFO: [y4mscaler] Input Stream Header:
INFO: [y4mscaler] <<< frame size: 1360x1020 pixels (4161600 bytes)
INFO: [y4mscaler] <<< chroma: 4:4:4 (no subsampling)
INFO: [y4mscaler] <<< frame rate: 25/1 fps (~25.000000)
INFO: [y4mscaler] <<< interlace: none/progressive
INFO: [y4mscaler] <<< sample aspect ratio: 1:1
INFO: [y4mscaler] Source matte region defaulting to full source frame.
INFO: [y4mscaler] Source active region defaulting to full source frame.
INFO: [y4mscaler] Target interlacing defaulting to match source.
INFO: [y4mscaler] Target active region defaulting to full target frame.
INFO: [y4mscaler] Deriving ratios from active regions and SARs...
INFO: [y4mscaler] ...using scaling ratios which pad target.
INFO: [y4mscaler] ...using scaling ratios which are simple.
++ WARN: [y4mscaler] Target active region clipped by projection of source.
INFO: [y4mscaler] === SOURCE parameters: =================
INFO: [y4mscaler] < stream:
INFO: [y4mscaler] < 1360x1020, SAR 1:1, progressive
INFO: [y4mscaler] < chroma subsampling: 444
INFO: [y4mscaler] < chroma ss ratios: x 1:1 y 1:1
INFO: [y4mscaler] < active region:
INFO: [y4mscaler] < 1360.000000x1017.659134 at 0.000000,1.222222
INFO: [y4mscaler] < matte region:
INFO: [y4mscaler] < 1360x1020 at 0,0 (bg Y'CbCr: 16,128,128)
INFO: [y4mscaler] === SCALING parameters: ================
INFO: [y4mscaler] | Scaler: Matto's Generic Scaler
INFO: [y4mscaler] | kernels: sinc:8, sinc:8
INFO: [y4mscaler] === TARGET parameters: =================
INFO: [y4mscaler] > stream:
INFO: [y4mscaler] > 720x576, SAR 59:54, progressive
INFO: [y4mscaler] < chroma subsampling: 420_MPEG2
INFO: [y4mscaler] < chroma ss ratios: x 1:2 y 1:2
INFO: [y4mscaler] > active region:
INFO: [y4mscaler] > 680x556 at 20,10 (bg Y'CbCr: 16,128,128)
INFO: [y4mscaler] > X ratio: 1/2
INFO: [y4mscaler] > Y ratio: 59/108
INFO: [y4mscaler] Output Stream Header:
INFO: [y4mscaler] >>> frame size: 720x576 pixels (622080 bytes)
INFO: [y4mscaler] >>> chroma: 4:2:0 MPEG-2 (horiz. cositing)
INFO: [y4mscaler] >>> frame rate: 25/1 fps (~25.000000)
INFO: [y4mscaler] >>> interlace: none/progressive
INFO: [y4mscaler] >>> sample aspect ratio: 59:54
INFO: [yuvcorrect] input: frame size: 720x576 pixels (622080 bytes)
INFO: [yuvcorrect] input: chroma: 4:2:0 MPEG-2 (horiz. cositing)
INFO: [yuvcorrect] input: frame rate: 25/1 fps (~25.000000)
INFO: [yuvcorrect] input: interlace: none/progressive
INFO: [yuvcorrect] input: sample aspect ratio: 59:54
INFO: [yuvcorrect] no line switching
INFO: [yuvcorrect] no time forwarding
INFO: [yuvcorrect] With luminance correction
INFO: [yuvcorrect] With chrominance correction
INFO: [yuvcorrect] Without rgb correction
INFO: [y4mscaler] Frame number 0
INFO: [y4mscaler] Frame number 1
INFO: [y4mscaler] End of stream at frame 1.
INFO: [yuvcorrect] Frame number 0
INFO: [yuvcorrect] End of stream!
INFO: [yuvcorrect] Normal exit: end of stream with frame number 1!
INFO: [mpeg2enc] SETTING AltiVec for MOTION!
INFO: [mpeg2enc] SETTING AltiVec for TRANSFORM!
INFO: [mpeg2enc] SETTING AltiVec for PREDICTION!
INFO: [mpeg2enc] Selecting DVD output profile
INFO: [mpeg2enc] Assuming norm PAL
INFO: [mpeg2enc] Progressive input - selecting progressive encoding.
INFO: [mpeg2enc] Encoding MPEG-2 video to xxx.m2v
INFO: [mpeg2enc] Horizontal size: 720 pel
INFO: [mpeg2enc] Vertical size: 576 pel
INFO: [mpeg2enc] Aspect ratio code: 2 = 4:3 display
INFO: [mpeg2enc] Frame rate code: 3 = 25.0 (PAL/SECAM VIDEO / converted FILM)
INFO: [mpeg2enc] Bitrate: 9800 KBit/s
INFO: [mpeg2enc] Quality factor: 1 (Quantisation = 1) (1=best, 31=worst)
INFO: [mpeg2enc] Field order for input: none/progressive
INFO: [mpeg2enc] Sequence unlimited length
INFO: [mpeg2enc] Search radius: 16
INFO: [mpeg2enc] GOP SIZE RANGE 9 TO 15
INFO: [mpeg2enc] Setting colour/gamma parameters to "PAL B/G"
INFO: [mpeg2enc] Progressive format frames = 1
INFO: [mpeg2enc] Using default unmodified quantization matrices
INFO: [mpeg2enc] Buffering 33 frames
INFO: [mpeg2enc] Signaling last frame = 0
INFO: [mpeg2enc] SETTING AltiVec for QUANTIZER!
INFO: [mpeg2enc] GOP start (1 frames)
INFO: [mpeg2enc] Frame end 0 I quant=9.04 total act=242.72241
INFO: [mpeg2enc] GOP start (0 frames)
INFO: [mpeg2enc] Guesstimated final muxed size = 84095
mpeg2dec-0.4.1-cvs - by Michel Lespinasse <[EMAIL PROTECTED]> and Aaron Holtzman
INFO: [pgmtoy4m] P5 cols: 720 rows: 864 maxval: 255
INFO: [y4mscaler] Input Stream Header:
INFO: [y4mscaler] <<< frame size: 720x576 pixels (622080 bytes)
INFO: [y4mscaler] <<< chroma: 4:2:0 MPEG-2 (horiz. cositing)
INFO: [y4mscaler] <<< frame rate: 25/1 fps (~25.000000)
INFO: [y4mscaler] <<< interlace: top-field-first
INFO: [y4mscaler] <<< sample aspect ratio: 59:54
INFO: [y4mscaler] Source matte region defaulting to full source frame.
INFO: [y4mscaler] Source active region defaulting to full source frame.
INFO: [y4mscaler] Target interlacing defaulting to match source.
INFO: [y4mscaler] Target frame size defaulting to match source.
INFO: [y4mscaler] Target active region defaulting to full target frame.
INFO: [y4mscaler] Deriving ratios from active regions...
INFO: [y4mscaler] === SOURCE parameters: =================
INFO: [y4mscaler] < stream:
INFO: [y4mscaler] < 720x576, SAR 59:54, top-field-first
INFO: [y4mscaler] < chroma subsampling: 420_MPEG2
INFO: [y4mscaler] < chroma ss ratios: x 1:2 y 1:2
INFO: [y4mscaler] < active region:
INFO: [y4mscaler] < 720.000000x576.000000 at 0.000000,0.000000
INFO: [y4mscaler] < matte region:
INFO: [y4mscaler] < 720x576 at 0,0 (bg Y'CbCr: 16,128,128)
INFO: [y4mscaler] === SCALING parameters: ================
INFO: [y4mscaler] | Scaler: Matto's Generic Scaler
INFO: [y4mscaler] | kernels: linear, linear
INFO: [y4mscaler] === TARGET parameters: =================
INFO: [y4mscaler] > stream:
INFO: [y4mscaler] > 720x576, SAR 59:54, top-field-first
INFO: [y4mscaler] < chroma subsampling: 444
INFO: [y4mscaler] < chroma ss ratios: x 1:1 y 1:1
INFO: [y4mscaler] > active region:
INFO: [y4mscaler] > 720x576 at 0,0 (bg Y'CbCr: 16,128,128)
INFO: [y4mscaler] > X ratio: 1/1
INFO: [y4mscaler] > Y ratio: 1/1
INFO: [y4mscaler] Output Stream Header:
INFO: [y4mscaler] >>> frame size: 720x576 pixels (1244160 bytes)
INFO: [y4mscaler] >>> chroma: INFO: [y4mtoppm] input stream parameters:
INFO: [y4mtoppm] <<< frame size: 720x576 pixels (1244160 bytes)
INFO: [y4mtoppm] <<< chroma: 4:4:4 (no subsampling)
INFO: [y4mtoppm] <<< frame rate: 25/1 fps (~25.000000)
INFO: [y4mtoppm] <<< interlace: top-field-first
INFO: [y4mtoppm] <<<sample aspect ratio: 59:54
4:4:4 (no subsampling)
INFO: [y4mscaler] >>> frame rate: 25/1 fps (~25.000000)
INFO: [y4mscaler] >>> interlace: top-field-first
INFO: [y4mscaler] >>> sample aspect ratio: 59:54
INFO: [y4mscaler] Frame number 0
1 frames decoded in 0.00 seconds (0.00 fps)
INFO: [y4mscaler] Frame number 1
INFO: [y4mscaler] End of stream at frame 1.
[LARRY:~/x] sms% exit
Script done on Thu Jul 22 19:52:10 2004
