JPEG Slide show and Aliasing

astorvick wrote on 6/21/2004, 4:02 PM
Hi eveyone,

Just finished my first slide show and after rendering it to DVD, I noticed horrible aliasing artifacts on lines when event panning the camera in the Y axis. Should I be using the normal MPEG output or should I be using the 24 frame with 3-2 pulldown, since the JPEG files are all progressive in nature.
How do I get rid of these aliasing problems?

-Thank you !

Comments

ClipMan wrote on 6/21/2004, 4:45 PM
...don't start with jpegs... they're already compressed ...use high quality image files like Tiff or Png ... otherwise, your show is toast...
astorvick wrote on 6/21/2004, 4:50 PM
Well, their camera only takes Jpegs. I don't know of any consumer cameras the are native PNG. The quality is really good and I understand the compression/recompress thing, but what does that have to with Aliasing?
I think they used the 5 mpixel setting, 2592x1944x24.
ClipMan wrote on 6/21/2004, 4:55 PM
... it's the ppi (pixels per inch) that counts... it's gonna end up 72 ppi ... you'll wanna start with images at 200 -300 ppi uncompressed... you won't get that from a camera, I agree .... you're always gonna have problems unless you get your stills from somewhere else....
Cheesehole wrote on 6/21/2004, 5:14 PM
5MP is far more than enough for DVD.

The easiest way to solve the problem you describe is by blurring the image. Try adding a quick-blur to the problematic images, or even the whole track, and see how the stability of the image improves on a TV screen. If Quick Blur doesn't do it for you, there is Gaussian blur... you could just use a slight vertical blur to specifically address the Y-motion. I ended up using the median filter, but it is very slow.

I would definitely render to progressive. But make sure you preview on a normal TV because you need to see how the details of the image work with the interlaced display. You might need to add more blur.

The artifacts could be a couple things... the MPEG compressor running out of bits on an image with too much fine detail, or interlace artifacts - also from having too much fine detail. Luckily bluring will solve both problems.
joelaff wrote on 6/21/2004, 8:02 PM
Make sure you have the Quality settings set to Best and not just good. Best will use Bicubic vs. Bilineaer interpolation for the scaling. Bilinear can lead to aliasing, especially when scaling.
johnmeyer wrote on 6/21/2004, 8:32 PM
JPEG is probably not the problem. Also, motion blur, while it will help it will (surprise) soften the image.

joelaff has the correct answer: Use Best for the render quality. This will make a big difference. It won't cure the problem 100%, but it will get close.

This has been discussed many many times. The bottom line is that Sony should change the names of the render options because "Best" is primarily for rendering still photos (search on Sony's name and the word render and you'll find the posts that verify this). "Best" works fine for video also, but it takes 2-3x longer, so it is a waste of time unless you are rendering stills. Sony should call the option "Still Photo" rendering, or something like that.

John_Cline wrote on 6/21/2004, 8:56 PM
Pixels per inch has absolutely nothing to do with video. It is all about vertical and horizontal resolution in pixels. Think about it for a moment, if you view a 720x480 video on a 12 inch TV and then a 21 inch TV, the two images will be very different in size. Also, if you have a 17 inch computer monitor set for a display resolution of 800x600, the 720x480 video will be a completely different size that if you viewed the image on the same 17 inch monitor set for 1600x1200.

Pixels per inch or dots per inch are only valid for printers. If you have a 1024x768 JPG, it will display exactly the same on the screen regardless of whether the image is 72 PPI or 300 PPI or 1000dpi.

John
johnmeyer wrote on 6/21/2004, 9:10 PM
Pixels per inch or dots per inch are only valid for printers.

John, your post is exactly correct.
ClipMan wrote on 6/22/2004, 5:24 AM
.... the jpeg lossy algorithm uses the surrounding pixels to resize the still ... this means a loss of colour, detail and definition that is particularly apparent on the horizontal and vertical edges of the still when you start manipulating it ... starting from a higher resolution gives the algorithm enough pixels to maintain better colour, detail and definition thereby making the edges appear less blocky and washed out... I've been using stills in video presentations for many years and there is no doubt that coming from a higher ppi source presents a better image in 2d or 3d space...
astorvick wrote on 6/22/2004, 7:32 AM
Thank you all for helping

I guess the only reason I gave the Mpixel setting was to give you the size of the image. Here is what I did to remove some of the aliasing.
I usually have my project setting at best, But I also rendered to 24P with 3-2 pulldown, that seemed to help the most, althought there is still a little when the it's panning in the Y axis/vertical, If I zoom the image I don't notice it that much. Again thank you for the help.

joelaff wrote on 6/22/2004, 9:46 AM
> usually have my project setting at best, But I also rendered to 24P with 3-2 pulldown, that seemed to help the most,

Probably the reason this helped is that , by lower the frame rate, you increase the motion blur for each frame. 24fps footage is considerably more blurry than 30p or 30i.

Anything I create myself I always do at 24p with 3:2 (which means Disable resampling, as with resampling on Vegas does not do a traditional 3:2 (3 whole frames and 2 split frames).
astorvick wrote on 6/22/2004, 12:24 PM
Anything I create myself I always do at 24p with 3:2 (which means Disable resampling, as with resampling on Vegas does not do a traditional 3:2 (3 whole frames and 2 split frames).

1. Are you saying that I should also disable resampling before I render still images?

2. Why doesn't the aliasing show on generated text media, isn't that square pixel and progressive?

Thank you so much for your help.
joelaff wrote on 6/22/2004, 1:09 PM
I may have said something confusing... If your project is at 24p and you tell Vegas to put in pulldown in the render then I am sure how that works. I meant if your timelin is 29.97 and you import 24fps footage. I have found that with resampling set to anything other than disable it will split most of the frames instead of the traditional 3whole/2split.

I am referring to imported footage being done at 24fps. If you use track motion I am not sure how you would mimic this. Probably you would have to render it to a 24fps clip and then import it onto a 29.97 timeline with resampling off.

But if you found something you like then great!