Video distortion - best way to correct

There is a free plug in made for the fisheye effect removal for Final Cut Pro X users available, so there's no need to use other sofware when editing your P2V videos with FCPX.

You can download it from here: http://www.alex4d.com/7-alex4d-wide-angle-fix-effect

Because my new P2V still hasn't arrived from Netherlands, I haven't been able to try this plug in for my own material, but I tried it with some P2V video material I downloaded from Vimeo and it seemed to be doing the fisheye removal quite nicely.

-jonzalez-
 
gfredrone said:
http://tv.adobe.com/watch/no-stupid-questions-with-colin-smith/photoshop-lens-correction-for-video/

Open in Ps
Convert to smart object
Lens Correct with DJI profile
Render lossless
Import to video editing application

Or you can shoot in a narrower FOV in the camera settings on the app.



Thanks dfredrone, did not know that, great info.
 
Or you can shoot in a narrower FOV in the camera settings on the app.

Unfortunately, you'll probably notice that the quality of the video goes down when you shrink the FOV. That's because the camera will be using less of the lens and sensor- as if it had an even smaller lens and sensor.

It's best to take video at the full FOV and process it "in post" to remove the distortion if you wish. You can use Photoshop on video as described above. I've also found that the Lens Correction effect in Adobe Premiere works pretty well. You have to scale the video or mask it, though, to hide the gaps left at the edge of the frame. When I get the time, I will produce a tutorial showing how I have done it specifically for the Vision. In the mean time, you can find lots of help on the web by searching for "fisheye removal" or whatever.

-- Roger
 
jadebox said:
Or you can shoot in a narrower FOV in the camera settings on the app.

Unfortunately, you'll probably notice that the quality of the video goes down when you shrink the FOV. That's because the camera will be using less of the lens and sensor- as if it had an even smaller lens and sensor.
I disagree. The video resolution is still 1920 x 1080, no matter what FOV you select. It's NOT cropping the video - it's sampling 1920 pixels from the sensor's full 4384-pixel width, and 1080 pixels from the 2922-pixel height. The FOV just changes where those pixels are sampled from.

You DO lose resolution if you crop it in post, however. And using less of the lens is the whole point of changing the FOV.
 
I'm using:

Virtualdub v1.10.4 (http://www.virtualdub.org/download.html)
Windows moviemaker (windows8)
VirtualdubFFMpegPlugin_0812_beta1

I found an good filter voor virtual dub to remove the fish-eye effect/barrel:
barrel distortion: emiliano ferrari (http://emiliano.deepabyss.org/)
I use the following settings:
ALPHA: -0.20200
BETA: 0.03500
deselect: aspect-ratio

All mentioned software is freeware for non-commercial use.
Example: http://youtu.be/-IXJkXRln1M

If it is the best way, I don't know.
 
jimre said:
jadebox said:
Or you can shoot in a narrower FOV in the camera settings on the app.
hi

Unfortunately, you'll probably notice that the quality of the video goes down when you shrink the FOV. That's because the camera will be using less of the lens and sensor- as if it had an even smaller lens and sensor.
I disagree. The video resolution is still 1920 x 1080, no matter what FOV you select. It's NOT cropping the video - it's sampling 1920 pixels from the sensor's full 4384-pixel width, and 1080 pixels from the 2922-pixel height. The FOV just changes where those pixels are sampled from.

You DO lose resolution if you crop it in post, however. And using less of the lens is the whole point of changing the FOV.

You lose quality, not resolution. The lens on the Vision is fixed. So, to reduce the FOV, the software in the camera samples fewer of the sensor's pixels to produce each pixel in the video. At the full FOV, the camera is using the full sensor and all of the lens. At a smaller FOV, it is using a smaller portion of the sensor and less of the light that passed through the lens.

At the full FOV, the camera is using all of the sensor's pixels to capture the image and sampling the 14MP image down to the video's resolution. To produce a smaller FOV, the camera is just looking at a subset of all the sensor's pixels - the ones near the middle. So, for each pixel of the video, it will use less, perhaps just one, of the sensor pixels. This introduces more noise into the image, reducing the quality.

Another way to think of it is that, since the lens is fixed, producing a smaller FOV image is the same as cropping the full image and scaling it up to the original resolution.

So, as I said, if you choose a smaller FOV, it is as if you are using a camera with a smaller sensor and lens.

Camera makers have caused a lot of confusion by suggesting that more megapixels equals a better photo. But, generally, a larger sensor and better lens will produce much better photos.

-- Roger
 
jadebox said:
At the full FOV, the camera is using all of the sensor's pixels to capture the image and sampling the 14MP image down to the video's resolution. To produce a smaller FOV, the camera is just looking at a subset of all the sensor's pixels - the ones near the middle. So, for each pixel of the video, it will use less, perhaps just one, of the sensor pixels. This introduces more noise into the image, reducing the quality.
As far as I know, the camera is simply using Row- and Column-skipping to pick 1920x1080 pixels from the full 4384x2922 sensor. There's NO WAY this cheap, underpowered camera is recording the full 3K video frame then "down-sampling" in real-time. So while it may use the whole sensor width in wide-FOV mode, it's NOT using all the pixels. It's only using every 2nd or 3rd pixel for video.
Another way to think of it is that, since the lens is fixed, producing a smaller FOV image is the same as cropping the full image and scaling it up to the original resolution.
Nope. That's my whole point. This is BETTER than cropping and re-scaling back up - since it's NEVER less than 1920x1080 resolution.

Your suggestion, on the other hand - that it's better to shoot the full width and then crop in software - will DEFINITELY lose significant resolution. To get a 90° FOV your 1920x1080 video will be cropped to around 1234x694.
 
jimre said:
jadebox said:
At the full FOV, the camera is using all of the sensor's pixels to capture the image and sampling the 14MP image down to the video's resolution. To produce a smaller FOV, the camera is just looking at a subset of all the sensor's pixels - the ones near the middle. So, for each pixel of the video, it will use less, perhaps just one, of the sensor pixels. This introduces more noise into the image, reducing the quality.
As far as I know, the camera is simply using Row- and Column-skipping to pick 1920x1080 pixels from the full 4384x2922 sensor. There's NO WAY this cheap, underpowered camera is recording the full 3K video frame then "down-sampling" in real-time. So while it may use the whole sensor width in wide-FOV mode, it's NOT using all the pixels. It's only using every 2nd or 3rd pixel for video.
Another way to think of it is that, since the lens is fixed, producing a smaller FOV image is the same as cropping the full image and scaling it up to the original resolution.
Nope. That's my whole point. This is BETTER than cropping and re-scaling back up - since it's NEVER less than 1920x1080 resolution.

Your suggestion, on the other hand - that it's better to shoot the full width and then crop in software - will DEFINITELY lose significant resolution. To get a 90° FOV your 1920x1080 video will be cropped to around 1234x694.

What I said is correct. Even if it weren't, there would be a loss in quality just from using less of the lens. It's like having a smaller lens.

The difference in quality is apparent just by looking at the video.

-- Roger
 
OK, let's look at actual video to see the obvious difference. I just shot two identical scenes of a rock wall from a stationary P2V. This is a screen capture from a Premiere Pro project that overlays both video clips, and scales/crops the 140°FOV shot to match the 90°FOV shot as exactly as possible.

On the left is YOUR method - shoot at full FOV then crop/scale in software to get a 90°FOV. On the right is MY method - set the camera to 90° FOV. Which do you think looks better?
 

Attachments

  • fovtest.JPG
    fovtest.JPG
    150.1 KB · Views: 721
jimre said:
OK, let's look at actual video to see the obvious difference. I just shot two identical scenes of a rock wall from a stationary P2V. This is a screen capture from a Premiere Pro project that overlays both video clips, and scales/crops the 140°FOV shot to match the 90°FOV shot as exactly as possible.

On the left is YOUR method - shoot at full FOV then crop/scale in software to get a 90°FOV. On the right is MY method - set the camera to 90° FOV. Which do you think looks better?

I never suggested what you are refering to as "my method." Using the Lens Correction effect doesn't reduce the FOV by very much so there is very little scaling of the image and, therefore, very little decrease in quality.

-- Roger
 
jadebox said:
I never suggested what you are refering to as "my method." Using the Lens Correction effect doesn't reduce the FOV by very much so there is very little scaling of the image and, therefore, very little decrease in quality.

-- Roger
It sounded like you were warning people not to use the 90°FOV in-camera option. My point is that the video quality with this option is just as good - if not better - than the 140° FOV option.

I agree if you WANT an ultra-wide-angle image, but just want straight lines - then by all means shoot at 140° FOV and apply your favorite Lens Distortion correction software. But if you just want straight lines - and don't care about having an ultra-wide-angle - then the in-camera 90°FOV may be your best choice for image quality.
 

Recent Posts

Members online

No members online now.

Forum statistics

Threads
143,086
Messages
1,467,528
Members
104,965
Latest member
Fimaj