Jump to content

Stunt Rally DVD Intro Restoration Completed


ZANTHERA
 Share

Recommended Posts

So after around 40-50 hours of work over a few weeks I have finally completed the restoration of the Stunt Rally intro as found on the DVD-ROM release.

 

There were multiple issues with the intro such as the colors being overly green with blacks looking especially green, plus the issue of encoding errors from the 25p video being converted to PAL DVD which outputs at 50i.

 

Other errors were seen as the video changing from progressive to interlaced, what I mean by this is that with deinterlacing turned off some sections of the video would play just fine, it would show the same frame twice but then it would suddenly switch to interlaced footage and you'd see the lines of an interlaced frame that wasn't deinterlaced. When this happened there would be a transitional frame which would cause the video to freeze for a single frame, this happened quite a few times near the start resulting in the audio becoming out of sync when corrected. Towards the end this issue is very apparent as the video begins to bounce up and down due to constantly changing from progressive to interlaced and back again.

 

A frame showing the black of Sid's car as looking very green, also note how the jungle in the background has a minor green glow to it.

bUvnZNe.png

 

The same frame after color correction and upscaling to 1440 x 1080, while the blacks are much better the green cannot be fully removed due to the blacks being rendered more green than the surroundings.

xOpIN1R.png

 

(below) A frame that is a transition between progressive and interlaced playback.

rY68O8c.png

 

The next frame which has jumped down by a pixel but is now playing as progressive and has no field issues.

wlFY3Kf.png
 

Although a seemingly crude method I used MPC media player to play the extracted intro which was a lossless rip and go frame by frame print screening the video. If opened and left alone MPC will display a video in its native resolution in the center of the screen so the frames that were pasted into GIMP2 were 768 x 576, 4:3 PAL DVD resolution.

 

Luckily there were a few sections of the intro in the "behind the scenes" video that were of superior quality than those in the intro file itself, including the close up scenes of Baron and Brad but without the interlacing and bouncing issues, so these files were used instead since less work had to be done to fix the frames.

 

I did look into the possibility of extracting the frames from the data of the video using ffmpeg but this only resulted in giving me 1800 images from the top field and 1800 from the bottom, plus these images were not separated and would have required manual separation taking weeks if done that way. Since MPC displays the video exactly as is with these same frames but already deinterlaced and identical quality as the manually extracted frames it was decided to simply use the video itself and go frame by frame print screening. The screen image was then pasted into a template already set at 768 x 576 so all I needed to do was line up the top corner of the video with the top corner of the template to get the frame perfectly centered every time.

 

A color correction profile was already set so all I had to do was import the screenshot, line up just the video frame, copy that into a 1440 x 1080 template, drag it to the bottom left corner and upscale it using the best setting I tested which was Sinc (lanczos3), then a quick brightness profile set, then the color correction profile. Once all that was done the images were exported and called image-000 to image-1800 since ffmpeg can make an image sequence out of images as long as they start with image-000.

 

After all the frames were finished the sequence was made into a video with ffmpeg and then the black boxes along the side and interlacing lines on the top and bottom were cropped out and the video set back to 1440 x 1080 with AviSynth. After this a test upload was done and it was determined that the video being 25fps caused YouTube to compress it much more than usual, so the video was made back into a sequence of images and then each frame was doubled and all the files batch renamed from image-000 to image-3599. So the final video ended up being presented the same way the DVD would have if there were no interlacing issues, and thanks to ffmpegs compression methods the file size was not bigger than the 25fps version.

 

The audio was next and for the longest time I was using the audio from the DVD file that I attempted to fix the eqalisation on and do some noise reduction due to hiss being very present. However in the last few days a higher quality version came to light when I was digging through the extracted RF files in the ART0001 folder, there were many audio tracks of superior quality found and in one of the folders appropriately named "intro" there was a file called "legointro" so I played it and amazingly it turns out to be the same audio used in the 320 x 240 intro but not cut down to a bit depth of 8, instead it was the almost original file being 22,050Hz PCM 16 bit audio. So hidden in the files the whole time was the exact file I needed for my restored video, using the same technique as they did with the DVD audio I mirrored the frequencies from 0Hz to 11,025Hz up past the previous quality peak and made the track 44,100Hz. After this much more work was done to the equaisation and I also separated the central, left and right channels and increased the outer channel amplification to give the whole track a more spacious sound, the frequencies above 16,000Hz were then deleted as the data present from here up caused very high frequency noise that did not work well with the lower frequencies.

 

Due to the duplicated frames the audio did have to be sped up by 0.75 percent so if it sounds a bit pitched up that is why, also there is a section towards the end after Brad and Baron jump the ice ramp that the video is still for just under a second, this is because in the next scene there are missing frames for some reason so duplicate frames had to be added to keep the audio in time with the video.

 

Here you can see the file after the frequency mirror was performed, the frequencies above 11,025Hz are duplicated from below and does increase the quality of the track if equalised properly.

 

lHEswAt.png

 

So here it is all done! A huge thank you to Mr. Skeltal for helping me with using ffmpeg, AviSynth and a few other programs which would not have made the restoration possible. Also a big thank you to Cyrem for his RF extraction tool which enabled me to find the higher quality intro music.

 

 

Here is the full quality version which came out to a decent 135 MB, uncompressed video at 15,000 kbps and full quality PCM 44,100Hz restored audio.

 

Link to comment
Share on other sites

  • 2 weeks later...
52 minutes ago, jojo337 said:

It sounds good ! But the video is deleted ??

 

*cough cough*

 

(And no, I don't know why it was re-uploaded nine days ago)

  • Like 1
Link to comment
Share on other sites

Sorry it was reuploaded because the first restoration of the audio had some bad equalisation on it so I decided to redo it, then I uploaded it forgetting that YouTube for some reason puts a delay on the audio so had to upload it yet again, I have now fixed the link in the first post and will update below that with a new MEGA link once it is uploaded.

Link to comment
Share on other sites

 Share

×
×
  • Create New...

Important Information

We have placed cookies on your device to help make this website better. You can adjust your cookie settings, otherwise we'll assume you're okay to continue.