Zoom : adding 'no aspect ratio change' in general settings (1 Viewer)
- Thread starter miljbee
- Start date
Greece
- Thread starter
- #33
here are one exemple of what smart zoom could produce :
Original image :
Zoomed 10% (10% of height is lost) :
Zoomed 10% + Smart Expand (sides are more expanded than center) :
Zoomed 10% + Smart Expand + 5% Brut width expand :
And here is how i performed smart expand :
The third picture doesn't look so bad, even if my smart expand is a bit too simple. instead of cutting the picture into 6 pieces, it could be cut into 20 or 50 to get a far better result and perhaps managing to fill in the whole screen.
Of course, some of you will argue that the choosen picture isn't a good choice because there isn't any interresting things to see to the right and left of the picture. Anyway, most of us are watching this type of content.
Propose me a picture with something to see on the sides, and i will try to do the same.
oh, and sorry for the poor quality of the pictures, i have done all this with paint. if someone is interrested to produce some better example built with better soft, don't hesitate ! perhaps even that some sort of "photoshop" could perform the smart expand in a real progressive manner.
Original image :
Zoomed 10% (10% of height is lost) :
Zoomed 10% + Smart Expand (sides are more expanded than center) :
Zoomed 10% + Smart Expand + 5% Brut width expand :
And here is how i performed smart expand :
The third picture doesn't look so bad, even if my smart expand is a bit too simple. instead of cutting the picture into 6 pieces, it could be cut into 20 or 50 to get a far better result and perhaps managing to fill in the whole screen.
Of course, some of you will argue that the choosen picture isn't a good choice because there isn't any interresting things to see to the right and left of the picture. Anyway, most of us are watching this type of content.
Propose me a picture with something to see on the sides, and i will try to do the same.
oh, and sorry for the poor quality of the pictures, i have done all this with paint. if someone is interrested to produce some better example built with better soft, don't hesitate ! perhaps even that some sort of "photoshop" could perform the smart expand in a real progressive manner.
If it could detect, and crop black bars on all sides, this would make a lot of difference for us with analog tuners. I have a PVR-500, and it has irregular black areas on every side, so that when using the current zoom-modes everything looks distorted. The black areas are of course overscan area tht is broadcasted, but it still looks crap. And I have allready tried to use ffdshow processing, but it crops to much, or to little as the overscan differs..
I imagine that this "black area" detection would have to be carried out within the DirectShow filters, or maybe even as a self-contained filter as it is working with pixel data directly.
Performance could be an issue unless someone can describe a really fast and robust algorithm. Perhaps run the algo only on channel change and aspect change (implying that the user is not happy with the current look), thereby reducing average load.
Is it enough to "sample" a few chosen blocks, or should every "black area" pixel be fed into this algo? Does it make sense to demand that a pixel is 0,0,0 to be black, or do we need some threshold to allow for analog noise etc?
The model would be something like:
█████████logo
â–ˆactive image area..â–ˆ
â–ˆactive image area..â–ˆ
â–ˆactive image area..â–ˆ
â–ˆactive image area..â–ˆ
â–ˆactive image area..â–ˆ
â–ˆactive image area..â–ˆ
████████████
â–ˆPotential text here.â–ˆ
A simple way would be finding every vertical and horisontal line containing 60% or 90% black pixels ("black" defined somewhere). From line 1 downwards find the first line that is not black, this is the up border. Repeat for bottom, left and right.
k
Performance could be an issue unless someone can describe a really fast and robust algorithm. Perhaps run the algo only on channel change and aspect change (implying that the user is not happy with the current look), thereby reducing average load.
Is it enough to "sample" a few chosen blocks, or should every "black area" pixel be fed into this algo? Does it make sense to demand that a pixel is 0,0,0 to be black, or do we need some threshold to allow for analog noise etc?
The model would be something like:
█████████logo
â–ˆactive image area..â–ˆ
â–ˆactive image area..â–ˆ
â–ˆactive image area..â–ˆ
â–ˆactive image area..â–ˆ
â–ˆactive image area..â–ˆ
â–ˆactive image area..â–ˆ
████████████
â–ˆPotential text here.â–ˆ
A simple way would be finding every vertical and horisontal line containing 60% or 90% black pixels ("black" defined somewhere). From line 1 downwards find the first line that is not black, this is the up border. Repeat for bottom, left and right.
k
HI. Since I have an analog card PVR-500, and analog cable I thought I should show you how the current zoom-modes work. As you can see by the following pictures, it's not really usefull for me atm. I'm currently experimentiong with the new zoom 14:9, and it looks promising, but the black bar at the right looks bad still...
Have a look: (all images are uploaded in full res png)
Original and Normal (they look the same)
Letterboxed 4:3
Zoom 14:9
Zoom
Stretch
Pan and Scan
And one 4:3-source original..
MAN; after all this work I see that this is all using 4:3 A/R, and I thought it looked worse. Then it hit me - my screen is a 40" 16:9
- it looks way worse in 16:9.
That thought brings me to :idea: ;
- Was the aspectratios, and zoom modes originaly ment for 4:3 :?:
Have a look: (all images are uploaded in full res png)
Original and Normal (they look the same)
Letterboxed 4:3
Zoom 14:9
Zoom
Stretch
Pan and Scan
And one 4:3-source original..
MAN; after all this work I see that this is all using 4:3 A/R, and I thought it looked worse. Then it hit me - my screen is a 40" 16:9
- it looks way worse in 16:9. That thought brings me to :idea: ;
- Was the aspectratios, and zoom modes originaly ment for 4:3 :?:
- Thread starter
- #39
Hey, there is quiet a lot of great ideas in this post !
if you don't mind, let me organize a bit all this !
At first, and as a start point, i think that actual zoom feature need to be customizable by the user.
There is several way to achieve this.
the simplest is to ask the user to chose from existing zoom modes those he most uses. Then map those zoom modes to the zoom key and add a zoom menu to TV and video osd to let the user choose from all available zoom modes. if the devs have a few time to code this, it would really be great.
for customizable zoom modes, things could be improved by enabling the user to define it's very own zoom modes. This might be a bit tricky to find the good parameters to present to the user. But if it's well thought, it will be a really powerfull feature.
Second point : defining a non-linear zoom feature. This feature would enable to do smart shrink/expand (see NLS post - botom of the second page)
Third point : adding an automatic black borders detection algo. This would enable to automatically get rid off them and eventually present the interresting part of the picture into the whole screen.
Fourth point: defining a smart zoom feature that would combine all the three points above. The goal of the smart zoom is to combine all the avilable zoom method (crop, smart expand, AR modification) to get a biggest picture, not too "out of shape".
Would you like me to create four new subjects in this forum to clarify all this ?
if you don't mind, let me organize a bit all this !
At first, and as a start point, i think that actual zoom feature need to be customizable by the user.
There is several way to achieve this.
the simplest is to ask the user to chose from existing zoom modes those he most uses. Then map those zoom modes to the zoom key and add a zoom menu to TV and video osd to let the user choose from all available zoom modes. if the devs have a few time to code this, it would really be great.
for customizable zoom modes, things could be improved by enabling the user to define it's very own zoom modes. This might be a bit tricky to find the good parameters to present to the user. But if it's well thought, it will be a really powerfull feature.
Second point : defining a non-linear zoom feature. This feature would enable to do smart shrink/expand (see NLS post - botom of the second page)
Third point : adding an automatic black borders detection algo. This would enable to automatically get rid off them and eventually present the interresting part of the picture into the whole screen.
Fourth point: defining a smart zoom feature that would combine all the three points above. The goal of the smart zoom is to combine all the avilable zoom method (crop, smart expand, AR modification) to get a biggest picture, not too "out of shape".
Would you like me to create four new subjects in this forum to clarify all this ?
Idea for config:
I remember in the early days of downloading people "shrinked" the original to minimize a movie to fit onto one CD-r. I used Zoomplayer, and it had a feature to "drag and drop" the picture onto a grid while playing. What if the user could do that in config with the sources (tv/video/dvd), and save it as a zoom-mode?
I remember in the early days of downloading people "shrinked" the original to minimize a movie to fit onto one CD-r. I used Zoomplayer, and it had a feature to "drag and drop" the picture onto a grid while playing. What if the user could do that in config with the sources (tv/video/dvd), and save it as a zoom-mode?