May 7, 2011

• Minami-ke! 5 no 2! Utamatsuri Dayo! Hokago Daibakuhatsu!! +1000 pix

  

  Today I'm sharing my DVD-rip of aforementioned live concert, which was released 2009.04.22. Well, it was kinda 2 years ago, and I came up with it only now, late you say? Not late until someone makes it before me, that's what I think. And the only 2 rips of this DVD I found so far were of quite low quality (1.3 Gb one) or low availability (2.5 Gb one, from Perfect Dark). In case you expect to see the performance review, go to danie's blog entry, otherwise enjoy the "making of" the rip here:

  Actually, the quality issue is what made me to create dvd-rip myself. Let's see what we have on the original, untouched DVD:

  1.  Low contrast (in fact, it seems to have incorrectly set TV/PC levels, that is, black is at 16 where it should be at 0 and white - 235 instead of 255)

   

  ^ What was on the DVD.

   

  ^ What I made it look like in the rip.

  2. Low vertical resolution.

  While previous issue is barely an issue at all, since it could be more or less easily fixed on playtime (by changing settings of the decoder), this one is more serious. Look at the pictures:

  

  While DVD and rip have the same vertical size, the DVD has twice as lower vertical resolution (the horizontal difference is okay, this is how a picture is stored on DVDs, since horizontal resolution is less important). This is especially noticeable on the skirt and line on the left. Such an issue is not usual for DVDs, and is barely if at all possible to fix on runtime. Seems like no one of rippers of this concert paid attention to it, since both rips I found had the same problem (some even less vertical size!). Then I decided to make one myself.

  3. Unbalanced sharpness distribution.  

  This is less noticeable, but very common for DVDs, since DVD makers tend to overblur the content to save space while sharpening it to regain visibility. Unfortunately this usually leads to mentioned above difference between "flat" areas which make 90% of image (and still have compression artefacts such as DCT blockiness, that appears due to low detail level) and oversharped edges with halos around them (especially noticeable in combination with low contrast on dark scenes around light sources). You can see what I'm talking about on the pictures above.

  So, how do I fixed all this stuff? 

  Avisynth.

  The average script I used while encoding is as follows:

   

  mpeg2source("video.d2v",cpu2="xoooxx")                   #had to use horizontal deblocking to control bitrate sometimes, usually it only increases blurriness.

  separatefields.selecteven                                #instead of automatic deinterlacing I had to use its manual alternative. By splitting each field into 2 frames and discarding every 2nd of them I got rid of useless blurry duplicates (they were the reason of low vertical resolution, since instead of being 2nd half of the image they were just smoothed out copy of its 1st half. You can get more info about what interlacing is here: 100fps.com)

  crop(4,2,-2,0)                                           #usual procedure of cropping black borders around image present on most DVDs

  bicubicresize(848,680)

  sangnom                                                  #here goes the magic. This procedure allows to recreate (or at least make it look like) lacking vertical resolution, but it needs to have extra vertical size on its input to avoid damaging the actual details.

  bicubicresize(848,480)                                      

  x=sharpen(0.6).blockbuster(method="noise")               #this fixes lack of sharpness and detail on the image. By adding noise to dark areas of the picture I make encoder pay more attention to it, and since concert videos are usually mostly dark, this allows to increase perceivable quality of the image (and greatly increase needed bitrate though).

  maska=mt_edge(mode="prewitt").mt_invert.mt_inpand.mt_deflate

  dehalo_alpha(ry=2.5)                                     #this fixes halos around oversharped edges of the image, especially vertically. Should be careful though, extra usage could make image look unnatural.

  overlay(x,mask=maska,opacity=0.6)                        #combining sharped details and lowered halos into one using edge mask

  levels(16,1,235,5,250)                                   #now it's time to change contrast to what it should be. The PC scale (0-255) looked too dark, so I decided to make it a bit closer to input.

  gradfun_db(1.3)                                          #fixing broken gradients which usually appear after increasing contrast, and further improving perceivable quality by adding subtle noise to them

  sharpen(0.3)                                             #a few final touches, make image look more detailed after expanding it from DVD's 3:2 to watching 16:9 (now I think I should've sharped it even more)

   

  And that's it, not a very difficult procedure, but it sure makes difference, just check out some of the existing rips and compare to mine.

  As for the encoding part, I used VFW version of the ultimate x264 codec (the one used in Virtual Dub), and split the source video into 4 parts. This allowed me to use variable bitrate mode (crf=~15.7) while maintaining final size to be exactly 1 DVD-R (4.4 Gb), by changing script settings for each part separately. (Well, to be honest, the resulting size was a bit larger than 4488 Mb, so I had to fix it by cutting out King Records logo animation at the end of each DVD. But people downloading DVD-rips don't care about recording companies anyway, right?)

   

   For those, who made it till here, I am sharing the photo of the best performance ever, Keikenchi Joshochuu by Minami-ke cast:

  
  (I could write about just as long "making of" for this very picture, so...) 
  Ah yes, the link to the rip. It is here: http://www.mediafire.com/?pwvdowsi4dvbf. So that you actually had to read it till here anyway.

  • 5:41 pm
  • 5 Comments