--------------------------
- Vários fatores devem ser considerados quando se captura vídeo de uma fonte VHS. Eu não sou nenhum especialista em vídeo digital, mas através do doloroso processo de tentativas escolhi uma resolução para minhas capturas e compartilharei minha experiência. Assim você não desperdiçará tanto tempo como eu capturando e comprimindo diversas vezes.
--------------------------
- Esteja ciente de que você deve possuir os direitos para fazer uma cópia de qualquer material registrado, como filmes ou clipes de vídeo. Se você não possuir os direitos autorais, prossiga por seu próprio risco. Falando sobre riscos, eu não me responsabilizarei pelo uso dos procedimentos descritos neste documento. Isto é somente um um guia para o ajudar com suas capturas. Qualquer dano causado a seu sistema pelos procedimentos aqui mencionados será problema seu, e não meu.
--------------------------
- Se você procurar informações sobre a resolução de VHS na internet, você achará basicamente 352x240 com 29.967 fps (Quadros Por Segundo) para NTSC e 352x288 com 25 fps para PAL-M. Se você capturar com esta resolução, o resultado será ruim. Sendo assim, será que todos esses documentos encontrados estavam errados? Não, eles estão inexatos, ou talvez um pouco incompletos. O vídeo VHS trabalha com o sistema entrelaçado (não progressivo), que significa desenhar cada quadro duas vezes na tela (um campo com as linhas pares e o seguinte com as linhas impares). Na realidade, todo quadro de um vídeo entrelaçado é dividido em dois campos (a razão para isto é uma história longa que vem dos primeiros TELEVISORES. Você pode achar mais informação sobre isto na seção de artigos de www.divx-digest.com). Devido ao entrelaçamento destes dois campos para formação do quadro, este possui muito mais do que 240 ou 288 linhas de informação. Se você captura às 352x240 ou 352x288 que você estará capturando apenas o primeiro campo e sempre perdendo o segundo, obtendo desta forma somente a metade da resolução.
--------------------------
- Outro ponto interessante é a compressão que você usa. Algoritmos de compressão (ex: xvid/Divx; -) ou indeo da intel ou o VCR1 e VCR2 da ATI, etc.) dividem a informação de vídeo em macroblocos de16x16 pixel. Quando menor a resolução que você usa, maior e mais aparente serão os macroblocos. Assim, até mesmo se sua fonte vídeo for de resolução muito baixa, você deve capturar com pelo menos 320x240, ou os resultados parecerão estranhos. Também, se você pretende recomprimir este vídeo posteriormente, ou se você pretender editá-lo, tente capturar com a menor compressão possível. Se você possui bastante espaço livre, e seu HD pode manter altas taxas de transferência, use formatos descomprimidos. Ao capturar um filme que tem legendas ou qualquer texto pequeno com a resolução padrão 352x240, você obterá pequenos " fantasmas " ao redor do texto após comprimi-lo com uma alta taxa de compressão, como xvid/Divx; -). Como uma foto jpg muito comprimida. O mesmo efeito (defeito) aparecerá nos contornos fortes (borda entre regiões contrastantes da imagem). Também há uma grande diferença entre vídeos capturados de uma fonte de VHS e os ripados de DVDs: o ruído. Se você comprime um vídeo ripado de DVD para a resolução de 352x240 com divx, você obtêm um resultado bom, porque a fonte do vídeo não possui ruídos. Por outro lado, se você comprimir um vídeo de fonte VHS com a mesma resolução, o ruído será visível e ampliado, e aumentará o bitrate necessário.
--------------------------
- Se você quer que seu vídeo capturado pareça-se igual a fonte de VHS (infelizmente não podemos ver algo melhor), capture com 640x480 a 29.967 fps, e então DESENTRELACE. Você pode fazer isso com softwares como Adobe Premiere ou VirtualDub (este ultimo é altamente recomendado - faça um download a partir da seção de software do www.divx-digest.com). Porém, esta resolução exige muito poder de processamento e espaço livre em disco. Alguns peritos diriam que tal resolução é muito alta para uma fonte de VHS sendo um desperdício de tempo e espaço, mas eu insisto: se você deseja perfeição, tente esta resolução. Agora, se você não precisa de tal perfeição, ou seu sistema não suporta, use 512x384 a 29.967 e o resultado será quase tão bom quanto 640x480. Eu tenho um ATI ALL IN WONDER PRO com 8Mb em um K6 2 400 MHz com 128 Mb RAM e esta resolução trabalha bem com o codec ATI VCR2. Por último, se você está sem espaço de disco, ou em um sistema lento, ou até mesmo se você quer fazer um vídeo compatível com VCD (que DEVE ter 352x240 com 29.967fps), capture com 352x480 com 29.967fps (deste modo você captura ambos os campos), e então redimensione para 352x240 com o VirtualDub usando o recurso de filtro " precise bicubic ". Este procedimento está descrito no home-page da Divx-Digest.
--------------------------
- A resolução de captura dependerá da qualidade final pretendida ou de sua necessidade. Capturando com a resolução padrão de VHS, todo segundo campo do vídeo não será capturado, resultando em perda de qualidade, desde que o sistema entrelaçado desenha "meio quadro " a 60 Hz, e não quadros realmente completos a 30 Hz. Capturando com a resolução de 512x384 com 29.967 fps ambos os campos serão capturados, dando um resultado bom e não exigindo muito poder do sistema. Ao capturar de uma fonte de VHS com resoluções maiores que a padrão (ex.: 512x384), é recomendado que se desentrelace o vídeo capturado para economizar espaço e melhorar a qualidade.
--------------------------
- -Se sua placa de captura não pode capturar em resoluções como 352x480 ou 512x384, use VirtualDub para fazer isto escolhendo a opção " set custom format " debaixo do " menu de vídeo ".
- se você esta usando o codec ATI VCR2 (somente usuários de placas ATI), o sistema pode travar quando recomprimir o vídeo ou até mesmo ao selecionar esta função. ATI não explica isto, e sugere atualizar o "ATI vídeo player" para uma versão mais nova, mas não funcionou para mim. A solução que eu encontrei foi fazer "downgrade" ( retornar a uma versão anterior ) do ATI vídeo player para versão 530005 (eu estava usando 5350002 antes). Eu realmente não posso comentar a diferença entre as versões 5xxxxx. Tudo que eu posso dizer é que as versões mais novas têm mais problemas.
-Se você estiver capturando um filme longo, você enfrentará o limite de 2Gb para os arquivos AVI (sim, quando a Microsoft criou AVI, eles pensaram que 2Gb seriam mais que suficientes). E mesmo se você salvar seus arquivos com o padrão AVI 2.0 (que ultrapassa esta barreira), você ainda enfrentará o limite de 4GB para o sistema de arquivos do Windows 95/98. Para resolver este problema você pode usar o Windows NT ou pode capturar com VirtualDub (como eu disse antes, este software é altamente recomendado). VirtualDub pode dividir o vídeo em arquivos de 2Gb ou arquivos menores em tempo real durante a captura. Interpretando estes arquivos como se fossem um único arquivo grande.
--------------------------
- Você pode enviar qualquer comentário ou sugestões a dgfig@terra.com.br
INTRODUCTION:
What do you think about increasing the quality of your DivX ;) movie-CDs while decreasing the file size? Even more: making the encoding time shorter!! A miracle? No. It is possible (sometimes) with a VirtualDub function that I forgot to mention in my first article (see "Best resolution for capturing from a VHS source", by myself). To be honest, I didn't forget to mention it. I just didn't know it existed. Well, living and learning! That "almost miraculous" resource is called "inverse telecine", and I'll be sharing this experience with you now.
LIMITATIONS:
This article is useful if you are working with a video source that uses the NTSC color system (thus, uses 29.97 frames per second), and that was originally made for the cinema (24 frames per second). Most of the movies you find in your "Blockbuster video store" are included in this category. This is completely useless if you are capturing a TV show, for instance.
I don't know how this applies to PAL movies, which work with 25 frames per second, because all my equipment is NTSC. If I find anything about it, I'll post it as soon as possible.
LEGAL NOTICE:
It's never too much to say that making copies (it doesn't matter the media) of any copyrighted material is illegal, unless you own the copyrights or have an authorization from the copyrights owner. This is specially important in this article, since it deals mostly with the capture of movies first intended to the big screen. Also, I'm not responsible for any problems or damages caused by the information hereby expressed. Proceed at your own risk. If you disagree with any of these, stop reading this guide now. Otherwise, good luck.
PROCEDURE:
Let me go straight to the point: the NTSC video uses 29.97 frames per second. The films intended for the big screen (movie theaters) use 24 frames per second. So, when you put a cinema film into a NTSC cassette, some conversion must be made. How is that made? That's simple: first, the film is slowed down by 0.1 percent, to be at 23.97 frames per second. Then, a new frame is inserted in every block of four frames, resulting in 29.97 frames per second. The problem begins in that "frame insertion". You can't just add a blank frame or repeat a frame, which would make the movie choppy. Thus, the video fields are used. At this point I suppose you know what video fields are. If you don't, this article is not YET for you. Try to learn about video fields and come back later. Here's what is actually done to every block of four frames:
-The first frame is kept untouched.
-The second converted frame is a mix of the first and the second original frames, both separated by the fields, which means that the even field is the first original frame and the odd field is the second original frame.
-The third converted frame is a mix (the same way as above) of the second and third original frames.
-The fourth converted frame is the third original frame.
-The fifth converted frame is the forth original frame.
At the end of all that, what you get is a sequence of frames that goes like this: three non-interlaced frames, then two interlaced frames, then three more non-interlaced frames followed by other two interlaced frames and so on.
Every interlaced frame will be a mix of two actual frames, and in case you have a high motion (a lot of difference between the two original frames that were mixed together), they will look like this:
INTERLACED VIDEO FRAME
Compressing an interlaced video source is a waste of space. Thus, something has to be done. Besides that, you would save 29.97 frames per second of a video that was made at 24 frames per second. That is 25% of completely useless data, because those extra 6 frames per second were not in the original movie!
If you just deinterlace your video before compressing (as I recommended in my previous article - sorry!), you are not getting rid of those extra frames. All you would do is mixing both fields together. The result would look like this:
DEINTERLACED VIDEO FRAME
As you can see from the picture above, both images are still present. I just got rid of the field lines. Besides, by doing this you lose half of your resolution, since the deinterlace process mixes the horizontal lines two by two.
And then there's this function called "inverse telecine", which can be found under the "video/frame rate" pull down menu in VirtualDub. What it does? When set to "reconstruct from fields - adaptive", it "demixes" the fields of the frames, so that they will be exactly how they originally were, eliminating one frame in every block of five, giving you the final result of 23.97 frames per second. The frames would then look like this:
INVERSED TELECINE VIDEO FRAME
Notice that no information of the original movie has been lost. It's all there as it should be, and you don't loose any resolution at all! Isn't this enough for you??? What if I say it's even faster than deinterlacing? (This sounds like one of those "Ginsu-Knife-Don't-Delay-Call-Today" TV show!).
To put a 2 hours long movie in a single CD using the simple deinterlace process, I used to use 600 kbits/second with low-motion DivX ;-). Now, after inverse telecine, I use 800 kbits/second and the movie fits the same space. That's 33% more quality! Notice that if you are using "inverse telecine", you don't need to (and you shouldn't) use the deinterlace filter. The "inverse telecine" does it all. Another important point is that you movie must have been captured at 29.97 fps and with both fields complete. That means 480 of horizontal lines.
SHORTENING IT ALL (THE PRACTICAL GUIDE):
-Capture at 512 x 480 @ 29.97 frames per second. VirtualDub does it, although I strongly recommend "freeVCR", which is a freeware that can be downloaded from "www.downloads.com" (perhaps you guys from divx-digest should try to put it on your downloads section - if it's not already).
-Load you captured movie into VirtualDub and set the resize filter to 512 x 384 with the "precise bicubic" option.
-Go to the "video" menu, then go to "frame rate" and choose "reconstruct from fields - adaptive" in the "inverse telecine" box. DO NOT change the frame rate. "inverse telecine" will do it automatically.
-For the compression, I recommend DivX :-) low motion @ 800 kbits/second if you want to put a 2 hour movie in a single CD.
-Choose your preferred audio compression (recommended is mpeg layer3).
-Set audio to "full processing mode".
-Go to the "audio" menu, then go to "interleaving". Set the "interleave audio every" to 30 frames.
-Save your work.
-Have fun.
Some further explanation about "inverse telecine" can be found in VirtualDub's help file. Enjoy!
What do you think about increasing the quality of your DivX ;) movie-CDs while decreasing the file size? Even more: making the encoding time shorter!! A miracle? No. It is possible (sometimes) with a VirtualDub function that I forgot to mention in my first article (see "Best resolution for capturing from a VHS source", by myself). To be honest, I didn't forget to mention it. I just didn't know it existed. Well, living and learning! That "almost miraculous" resource is called "inverse telecine", and I'll be sharing this experience with you now.
LIMITATIONS:
This article is useful if you are working with a video source that uses the NTSC color system (thus, uses 29.97 frames per second), and that was originally made for the cinema (24 frames per second). Most of the movies you find in your "Blockbuster video store" are included in this category. This is completely useless if you are capturing a TV show, for instance.
I don't know how this applies to PAL movies, which work with 25 frames per second, because all my equipment is NTSC. If I find anything about it, I'll post it as soon as possible.
LEGAL NOTICE:
It's never too much to say that making copies (it doesn't matter the media) of any copyrighted material is illegal, unless you own the copyrights or have an authorization from the copyrights owner. This is specially important in this article, since it deals mostly with the capture of movies first intended to the big screen. Also, I'm not responsible for any problems or damages caused by the information hereby expressed. Proceed at your own risk. If you disagree with any of these, stop reading this guide now. Otherwise, good luck.
PROCEDURE:
Let me go straight to the point: the NTSC video uses 29.97 frames per second. The films intended for the big screen (movie theaters) use 24 frames per second. So, when you put a cinema film into a NTSC cassette, some conversion must be made. How is that made? That's simple: first, the film is slowed down by 0.1 percent, to be at 23.97 frames per second. Then, a new frame is inserted in every block of four frames, resulting in 29.97 frames per second. The problem begins in that "frame insertion". You can't just add a blank frame or repeat a frame, which would make the movie choppy. Thus, the video fields are used. At this point I suppose you know what video fields are. If you don't, this article is not YET for you. Try to learn about video fields and come back later. Here's what is actually done to every block of four frames:
-The first frame is kept untouched.
-The second converted frame is a mix of the first and the second original frames, both separated by the fields, which means that the even field is the first original frame and the odd field is the second original frame.
-The third converted frame is a mix (the same way as above) of the second and third original frames.
-The fourth converted frame is the third original frame.
-The fifth converted frame is the forth original frame.
At the end of all that, what you get is a sequence of frames that goes like this: three non-interlaced frames, then two interlaced frames, then three more non-interlaced frames followed by other two interlaced frames and so on.
Every interlaced frame will be a mix of two actual frames, and in case you have a high motion (a lot of difference between the two original frames that were mixed together), they will look like this:
INTERLACED VIDEO FRAME
Compressing an interlaced video source is a waste of space. Thus, something has to be done. Besides that, you would save 29.97 frames per second of a video that was made at 24 frames per second. That is 25% of completely useless data, because those extra 6 frames per second were not in the original movie!
If you just deinterlace your video before compressing (as I recommended in my previous article - sorry!), you are not getting rid of those extra frames. All you would do is mixing both fields together. The result would look like this:
DEINTERLACED VIDEO FRAME
As you can see from the picture above, both images are still present. I just got rid of the field lines. Besides, by doing this you lose half of your resolution, since the deinterlace process mixes the horizontal lines two by two.
And then there's this function called "inverse telecine", which can be found under the "video/frame rate" pull down menu in VirtualDub. What it does? When set to "reconstruct from fields - adaptive", it "demixes" the fields of the frames, so that they will be exactly how they originally were, eliminating one frame in every block of five, giving you the final result of 23.97 frames per second. The frames would then look like this:
INVERSED TELECINE VIDEO FRAME
Notice that no information of the original movie has been lost. It's all there as it should be, and you don't loose any resolution at all! Isn't this enough for you??? What if I say it's even faster than deinterlacing? (This sounds like one of those "Ginsu-Knife-Don't-Delay-Call-Today" TV show!).
To put a 2 hours long movie in a single CD using the simple deinterlace process, I used to use 600 kbits/second with low-motion DivX ;-). Now, after inverse telecine, I use 800 kbits/second and the movie fits the same space. That's 33% more quality! Notice that if you are using "inverse telecine", you don't need to (and you shouldn't) use the deinterlace filter. The "inverse telecine" does it all. Another important point is that you movie must have been captured at 29.97 fps and with both fields complete. That means 480 of horizontal lines.
SHORTENING IT ALL (THE PRACTICAL GUIDE):
-Capture at 512 x 480 @ 29.97 frames per second. VirtualDub does it, although I strongly recommend "freeVCR", which is a freeware that can be downloaded from "www.downloads.com" (perhaps you guys from divx-digest should try to put it on your downloads section - if it's not already).
-Load you captured movie into VirtualDub and set the resize filter to 512 x 384 with the "precise bicubic" option.
-Go to the "video" menu, then go to "frame rate" and choose "reconstruct from fields - adaptive" in the "inverse telecine" box. DO NOT change the frame rate. "inverse telecine" will do it automatically.
-For the compression, I recommend DivX :-) low motion @ 800 kbits/second if you want to put a 2 hour movie in a single CD.
-Choose your preferred audio compression (recommended is mpeg layer3).
-Set audio to "full processing mode".
-Go to the "audio" menu, then go to "interleaving". Set the "interleave audio every" to 30 frames.
-Save your work.
-Have fun.
Some further explanation about "inverse telecine" can be found in VirtualDub's help file. Enjoy!
