| Author | Message | ||
     Michael Martin Rating: N/A |
hi folks, since i finally managed to buy 2 anaglyph movies on DVD (coming at ya, the bubble), i'd like to convert them to over/under and watch them with shutter/pageflip using M.H.'s great movie viewer. The DVD's come with Red/Blue cardboard glasses. I'll therefore write a filter converting each frame as an anaglyph picture to 2 (L/R) pictures. But how exactly do i convert every pixel? RGB -> RG- for left pixel RGB -> -GB for right pixel that is leave out Red for right picture and blue for left picture? or should i just use the red/Blue part as a gray scale pixel ? Is there a www description out there about red/blue, red/green, red/cyan and how to convert? thanks for your help | ||
| M.H. Rating: N/A |
It will be probably realy horrible work. DVD use YUV 4:2:0 quantization, it means that the R G B color will be mixed together. You will get in reality only 1/4 of the source resolution in stereo becouse of the color quantization ... Another factor witch must be taken in mind - human eye have diferent sensitivity for colors . It means that you can not easy convert R to white 1:1 . The sensitivity is 0.299R 0.587G 0.114B (maximum sensitivity for green) ... You must make an revers operation to get correct gray values for anaglyph. I am afraid that you will be forced to experiment a bit ... | ||
| Anonymous Rating: N/A |
You can try the batch conversion of 3dcombine, try both the grey scale and the colour fix, which tried to restore the colours (more or less successfully). Look at the reviews at www.stereovision.net for the link and similar software. You might need to extract the frames of the DVD movie (many GB needed) first. DVD software from http://go.to/dvdsoft or http://dvdsoft.free.fr | ||
| Anonymous Rating: N/A |
Where did you buy anaglyph DVDs? | ||
| M.H. Rating: N/A |
You shuld theoreticaly use folowing math: Forget about green. Devide blue by 2.6 to be in the same intensity as red for gray. Multiply both blue and red by 2 or so not to be in totaly black area. The result result of this multiplication shuld give the left and right eye inensitis in gray ... To get realy good results - the correct multiplications shuld be calulated after whole movie sequence histogram anaylysis ... Optimal method will be to use the mathematic revers of the algoritm with have cripled the movie from original to the anaglyph ... Could you place somwhere some sample sceenshots from the movies ? I could ty to derivate the correct equations from them .... | ||
| Michael Martin Rating: N/A |
Anonymous: I'm trying to use 'flaskMpeg' to render to avi (huffyuv) and 'virtualdub' to filter/convert. There is an API where one can write his own conversion very easy (from RGB buffer to RGB buffer frame by frame). From there i'll probably try to compress the over/under to MPEG2 again. Anonymous: I ordered them fom amazon.com http://www.amazon.com/exec/obidos/ASIN/6305564345/107-9820793-1399767, and http://www.amazon.com/exec/obidos/ASIN/6305538328/107-9820793-1399767 MH: Yes I'll experiment a bit, using your equation. meanwhile i've placed some frames on my webspace (as TGA without lossy comression and as jpeg) http://home.t-online.de/home/mmartin/anaglyph So probably different factors are needed for both films? how do i build a 'whole movie sequence histogramm'? is that a colour histogramm build over all frames ? could i automate the process of the factor derivation for my 'filter'? Another point is that i might want to adjust the 'plane of zero parallax', cause somteimes very eyestraining 'out of the screen' effects are used (at least in comming at ya). But that couldn't probably be done automaticly, as this plane also often changes during a scene, while the camera zooms in or out By the way, Coming at Ya has a '3D Explanation trailer' where a guy wearing polarized glasses (ore just shades?) states the film was taken using 'Dimensionscope 3D', whereas 'The bubble' has a label on the cover saying 'filmed in Space Vision' | ||
| M.H. Rating: N/A |
I mean calculate hstogram not only for individual frames , but for all frames in one shot. Or at least for several surronding images in the movie. You shuld obtain the range of gray with you could use for scaling the anaglypg conversion ... To do such calulation will require dual pass of the conversion ... It could be probably done automaticaly but it will require a lot of coding ... I have never heard about anybody who had encoded this ... I will try to have a look on the TGA images and let you know if I got any idea for good conversion math ... Even the automatic paralax correction could be thoreticaly encoded. The images could be fited one on another to minimalize the eye disparity by analysis of coeficient of its Fourier transformation (even vertical paralax and camera rotation colud be correctd in this way). I use such algoritms for electron microscopy images alignment. I do not unfortunately know about any implementation for movie usage. It is realy not trivial  . | ||
| M.H. Rating: N/A |
I was looking on the anaglyphs on your WWW ... Colud you, please, zip the TGA images ? They looks corrupted after downloading and I am not sure what is the reason ... I was trying to make some analysis at least of the JPG images .... After decomposition to R G and B chanels, it is clear that the math suggested by my will not work ... The Red chanel contains from some unknowen reason both left and right view ... Maybe I can get some better idea from the TGA images .. It is realy a pitty that the movies are not available on DVD in the form of interlaced stereoscopic movies in the same format as the stereoscopic VHS cassetes ... Non progresive mode of mepg2 compresion fully compatible with DVD colud be used with no compression ghosting for this purpose + standrard TV LCD shuterglasses ... | ||
| Michael Martin Rating: N/A |
let me see, if i've got this right: i calculate the histogramm over several (all) frames, divide the histogramm values for red by blue, and use that factor to multiply the blue value to yield grey (plus a 'brighness' factor for both values). To do that in one pass, i could deferr the conversion by some frames, and keep up an adapting histogramm to calculate the correction factor?. The range of gray i use in the output is only 256 gray values (equal values for RGB). Or is there another gray ramp possible with RGB?. As for paralax adaption, that sounds complicated and time consuming, doing a fourier transformation for every frame. Also that automatic would place the zero paralax plane to where most of the objects are (or the biggest), but not at the nearmost object (keeping everything inside the screen)? | ||
| M.H. Rating: N/A |
Yes, you get what I mean ... Unfortunately there is much bigger problem there - I am not able to find the correct equation for color to left/right transfer. The mixed left/right view in red chanel is something horrible. It is even visible in the anaglyph glasses ... Colud you plase put somwhwere some raw MPEG2 stream from the DVD (directly demultiplexed from the VOB file) to get realy acces to the raw data ? You are right according the alignment - it will only place the bigist equal shape areas on each other . maybe tha there will be any other faster agoritm availabe than the Fourier coeficient analysisi ... The best method will be manual anyway ... Do saome manual aligment for the first and last frame in each shot and interpolate the parallax corection between this point ... It sounds nice in theorey, I have never tried it in praxis . | ||
| mmartin@t-online.de Rating: N/A |
i would like to upload some raw mpeg2 data, but i don't know (yet) how to demux and cut a vob file to mpeg2 . by the way, both DVD's are not CSS encoded. | ||
| M.H. Rating: N/A |
Try to have a look on www.dvd.da.ru .... The demultiplexing could be done by th help of toll coled VobSnoopi. Unfortunately it can not cut the VOB to pieces + this tools have some bugs and works only for VOB files with correct beging ... Do you have some method how to cut a part from the VOB (with correct end and beging) ? | ||
| Michael Martin Rating: N/A |
using vobsplit and vobrator, i've uploaded a small sequence as cay.vob (vobsplit), cay-audio.vob (vobrator) and cay.m2v (demux with vobrator). finally i convertet also to cay.avi (flaskMpeg with indeo) which will be the source of my anaglyph conversion, but encoded with huffyuv and in full size..... | ||
| M.H. Rating: N/A |
Michael: I was trying to do evrything what I can = quality losless decompresion of the raw mpeg2 data + color histogram analysis of the results. I am not able to find any way how to conver this data to 2 separated images. The pictures do not look like anaglyphs, I wonder how it is possible to watch the movie by anaglyph glasses in stereo. It must be something horrible and the 3D effect are probably of low quality ... | ||
| Greg K Rating: N/A |
Regarding the "Comin' At Ya!" DVD, it works better than VHS, due DVD's component video format. But.. and here's the big let-down.. They tried to encode the film in color-anaglyph, meaning they tried to keep the original color information intact. Besides adding more ghosting, color anaglyph adds many unwanted side effects. Just look at a blue sky - color anaglyph encoded. The blue filter turns the blue sky white, while the red filter turns it black! This disc would have looked better if they would have stripped the original color info before encoding it to anaglyph. And as already mentioned, the best 3-D video method available is the field-sequential 3-D format, even with it's flicker. (A tad worse ala PAL) I own an old 3DTV copy of "Comin' At Ya!" in NTSC field-sequential 3-D, and it beats the anaglyph DVD version hands down. | ||
| M.H. Rating: N/A |
The good method will be to digitalize the original NTSC field-sequential VHS and convert it to above/below format for computer playback ... The best method will be to do this with the original film ... | ||
| Brightland Rating: N/A |
Hi, There is an alternative to shutter glasses and color anaglyph for video: Akumira. We recently finished converting a project from field-sequential stereo3D to Akumira. We then converted the footage to MPEG2, and played it on a laptop via composite video connected to a 50" HTDV (Mitsubishi: which upsamples the NTSC video to HDTV). Much to our surprise, it works reasonably well. Quality will be even better when played on a DVD player and connected via S-Video or component video. When played directly on a PC/laptop (via RGB), it looks excellent. It is not necessary to convert to grayscale to have quality color-filtered 3D: Akumira provides excellent and accurate color perception using red/cyan filters. While Akumira uses red/cyan filters, we do not refer to it as "color anaglyph" as it is impossible to get comfortable and accurate color perception with "color anaglyph". Akumira can display any full-color stereo3D footage with excellent color perception and comfort. We'll provide more info regarding products using Akumira in the near future. Regards, John | ||
| Giorgio Bogoni Rating: N/A |
Are you going to sell an Akumira-based plug-in filter for Premiere? If it would work as it works in R4D I think lot of us may be interested. |
