September 22, 2020, 03:54:37 am



Why Save as Image only at 72 dpi?

Started by aviphil87, December 03, 2015, 03:57:58 pm

Previous topic - Next topic


Dear all,

In 2.6.10....470 I try to take snapshots in a 1280x720 .mp4 video, taking advantage of Avidemux's capability to go frame-by-frame.

In the Avidemux window, slanted lines are just that, slanted lines, and contours are sharp. If I magnify the image with View/Zoom 2:1, still sharp.

When I Save as Image, i.e. as JPEG, and open the image in Photoshop (PS), slanted lines and contours become "stairs". When looking image info, it says 72 dpi, that's certainly where the stairs come from.

I don't see an option to get a JPEG with a better definition.

Strangely, if I copy the screen to the clipboard using the Print Screen key, and paste the clipboard in PS, the definition is still 72 dpi, and contours are stairs as well.

Is there any way to pick the actual video image, and get it in PS without a definition loss? Is it a Windows limitation? (Win7-32 in my case)

The image displays fine in the Avidemux window, so why not get it as-is in PS?

Thank you very so much for such a nice software, I don't dare to imagine how many man-hours in it...


December 04, 2015, 12:50:39 pm #1 Last Edit: December 04, 2015, 01:04:05 pm by AQUAR
The snapshot you take is at the pixel resolution of the video frame.
The 72 DPI is an arbitrary value for how those pixels are printed on paper.
Also the old analogue monitors where said to have a typical resolution of 72 DPI (was of course bit generalised!).
Meaning what you see on these old monitors is roughly what you get when printing.

Low resolution video (eg SD video) will have those staircase issues when you blow them up to the full screen of a typical HD PC monitor. The fix is to properly upscale such single frame pictures.
Nothing to do with Avidemux as its not a picture processing program.

I am sure there must be a resampling feature in photoshop.

Jan Gruuthuse

December 04, 2015, 03:57:25 pm #2 Last Edit: December 04, 2015, 03:59:25 pm by Jan Gruuthuse
not certain if this will make a difference:
Load video, select a frame, mark [A ] move one step further with right arrow and mark with [ B]
now use in avidemux Menu -> File -> Save as Image -> Save as JPEG
include frame saved as described above, from 3sat HD 720p recording.


December 05, 2015, 12:12:09 pm #3 Last Edit: December 06, 2015, 12:10:44 pm by AQUAR
@ Jan
Based on your sample, that also saves the frame at its native pixel resolution (pixels in the frame).
If the OP loads a 1080P video, then the jpeg resolution will be 1920 x 1080 pixels.
At that resolution there is no need for scaling the jpeg.
Meaning sloping edges will look the same, be it in the avidemux display or the jpeg in a picture viewer.

The 72 DPI is a bit misleading as it has nothing to do with the resolution of the jpeg picture.
You can change that value to say 300 DPI and the picture would still have the same pixel resolution (and look the same on a PC monitor!).

In the OP's case, comparing the captured frame with the rendered frame in ADM is like comparing native resolution with an upscaled version.

I normally use Spline interpolation to upscale low res pictures so as "maintain" smooth curved edges.


Folks, thank you for your input (what's OP, doc?). I know nothing about graphics display in a PC. So I tried to find info on the Web to make my mind and not tell stupid things, but I found nothing related to the suject matter.

And so, my question: In the 'Save as Image/JPEG' code, would it be difficult to set the JPG definition to 300 dpi instead of 72?

Perhaps at first glance it looks like a simpleton's question, but let me state it as follows.

Whichever processor, RAM, GPU et al on my PC, on my 1920x1080 screen, the Avidemux window shows a video frame on which a slanted line is a sharp slanted line. I.e., if I magnify the image with the largest View/Zoom ratio, the slanted line is still sharp.

Whichever internal code displays the video frame in the Avidemux window, the resulting picture is a raster, zillions of pixels just next to each other. And those zillions are such arranged that what they show is a slanted line which appears sharp.

But when these zillions pixels are either translated into a JPG, or are copied into the clipboard, the slanted line becomes a staircase.

In Photoshop or Gimp, when either opening the JPG, or pasting the clipboard, the image info shows that the definition is 72 dpi.

So it appears logic to think that the picture in the Avidemux window is at a higher definition than 72 dpi. Perhaps it is 300 dpi. Or if "dpi" means nothing here, let's just say that the definition of the Avidemux picture, looks like 300 dpi in Photoshop parlance.

It's frustrating to see the nice Avidemux picture with all its HD-video details, and not being able to get these details in PS or Gimp because of the low definition of the "transfering container".

Pardon me if the above reasoning is absurd. Thank you for your time and your patience,

Jan Gruuthuse

December 06, 2015, 05:06:23 pm #5 Last Edit: December 06, 2015, 05:15:24 pm by Jan Gruuthuse
OP = Original Poster
Still not certain I can follow your reasoning?
- Don't use avidemux as videoplayer: use VLC.
- Don't zoom in as you do now. If you need jpg screen capture from recording, use method as I tested and scale up later in Gimp.
Here you can find a scaled up picture as done with the example Posted before, you did click on thumbnail to view at full resolution?
1980 instead of 720:

If I'm not mistaken dpi is for paper output?
more details: The Simple Guide to Pixels, Resolution and dpi

as you can see the 1st image at 300 dpi instead of 72 dpi doesn't make any difference:
check in gimp if dpi is set to 300 dpi


@ aviphil87
Both Jan and Myself are trying to point out that a JPEG picture is comprised of a specific number of pixels.
If you print these at 300 DPI you get a small "tight" picture on paper.
If you print them at 72 DPI you get a larger "loose" picture on paper.

The DPI parameter has nothing to do with the resolution of the JPEG picture, it is a paper printer parameter.
The picture will be rendered at the DPI of the monitor regardless of that DPI parameter.
Up scaling will give the picture more pixels, and if done correctly will give a bigger smooth picture on your monitor. 
Magnifying (zooming) the picture, without up scaling, will give jagged edges.

When you take a snapshot of the frame AND want it to be smooth, it has to be up scaled.
There are features in every photographic program that lets you do that.

You are looking at this issue with an expectation that avidemux will upscale the frame capture for you.
AFAIK it doesn't do that - its a video editor that lets you save a selected frame (with a given number of pixels), as a jpeg picture.

Jan Gruuthuse

guessing mode: is your video
- 720p HD (Progressive)
- ###i (Interlaced)
QuoteInterlaced video creates a number of possible issues during capture, editing, and playback. The advantage of having a larger physical image with slightly more detail than you would find with 720p should be taken into account. Youââ,¬â,,¢ll get a larger, sharper image with 1080i than you will with 720p, but motion lines and other interlacing artifacts can certainly cause a disappointing overall viewing experience.

The display is it using
- digital port (HDMI,  Display port (Not analogue), ...)
- vga connector (could produce issues)
if the display also needs up scaling / down scaling this could also produce issues (HD Ready monitor with certain resolutions)


December 08, 2015, 10:10:35 am #8 Last Edit: December 08, 2015, 11:54:12 am by AQUAR
To me it looks like the OP is thinking of this issue in the same way as if using a flatbed scanner to scan (digitise) a photograph.
Here the DPI (dots per inch) of scanning the photograph equates to the sample size (number of pixels per inch "PPI") for digitising the photograph.

A video frame is already a digital picture with a particular pixel resolution (ie not PPI!).
Up/down sampling is used to render the video frame in a "windows box" of a particular size (number of horizontal pixels by number of vertical pixels).
I am not sure if that scaling is done by the rendering software or by the graphics card hardware.
Probably can be either/both (not ever thought much about this).
I am sure Mean knows the process intimately.

It's hard to convey the difference, but maybe someone else can be more succinct.