Discerning video quality

[Angel_Cyclops]

Hi, I dont know if this is the place to ask, but I have a question thats been bugging me for months.

There are 2 files of the same video both in 1080p, but one of them was upscaled from 720p(or 480p) to 1080p while the other was downloaded directly in 1080p. The thing to note is the upscaled video was encoded to match the original 1080p closely including the codecs/containers that were used, the fps, bitrate, audio sampling rate etc.

Apart from visual determination, is it technically possible to establish if a video file is transcoded and upscaled (or, that the original 1080p was not)?

[rrats]

No. mediainfo cant tell much about the video. you have to use your own eyes

[Angel_Cyclops]

No. mediainfo cant tell much about the video. you have to use your own eyes

Many times the upscaling is done crudely which skewers the video. Those are easy to tell. But often, with pros, they are done efficiently and are impossible to visually tell them apart.

[victoriabears]

An observation, I use roku and plex to play files, I have a non smart Sony TV which only goes to 1080, I find some files provided in 1080 do not look as good as 720, therefore can only conclude the replaying gear, roku,plex sony TV upscaling is doing a better job than the encoder did to 1080.

[rrats]

If I got you correctly you are trying to see if a video was upscaled without looking at it


If you are in fact looking at it then it's easy to tell. Upscaled video always looks lot uncanny

[Angel_Cyclops]

If I got you correctly you are trying to see if a video was upscaled without looking at it


If you are in fact looking at it then it's easy to tell. Upscaled video always looks lot uncanny

Since results from a visual examination can vary greatly due to its subjective nature, Im looking for a way to irrefutably demarcate between a transcoded video and one that isnt, and I think the only reliable way is if it can be compared through statistical values.

[johns0]

you can't get any quality comparison through statistical values,there's nothing in readings that will tell you a video was 1080p to begin with and that its a really good quality or that a 720 was upscaled.

[poisondeathray]

If I got you correctly you are trying to see if a video was upscaled without looking at it


If you are in fact looking at it then it's easy to tell. Upscaled video always looks lot uncanny

Since results from a visual examination can vary greatly due to its subjective nature, Im looking for a way to irrefutably demarcate between a transcoded video and one that isnt, and I think the only reliable way is if it can be compared through statistical values.

If you know what to look for, you can always tell them apart. The biggest difference will be high frequency details, those are the first to go. FACT: It's impossible to downscale and upscale without losing quality. Also that version incurs additional generation loss on top of the scaling (even if you didn't scale, just re-encode at the same bitrate, it would deteriorate, because you're not using lossless encoding)

Some cases might be harder to tell a) the original source was low quality, soft, to begin with (e.g. maybe it was an upscale to begin with) - ie. not really effectively 1080 lines (or close to) of real resolution; or b) the downscale step was a small amount, maybe 1-10% . c) the distribution bitrate was very high. The higher, the more difficult to tell differneces . And if a,b or c are correct and you cannot tell with your eyes, why do you care ?

If you have the known original you can use objective metrics, such as PSNR . This provides irrefutable proof.

[Angel_Cyclops]

If I got you correctly you are trying to see if a video was upscaled without looking at it


If you are in fact looking at it then it's easy to tell. Upscaled video always looks lot uncanny

Since results from a visual examination can vary greatly due to its subjective nature, Im looking for a way to irrefutably demarcate between a transcoded video and one that isnt, and I think the only reliable way is if it can be compared through statistical values.

If you know what to look for, you can always tell them apart. The biggest difference will be high frequency details, those are the first to go. FACT: It's impossible to downscale and upscale without losing quality. Also that version incurs additional generation loss on top of the scaling (even if you didn't scale, just re-encode at the same bitrate, it would deteriorate, because you're not using lossless encoding)

Some cases might be harder to tell a) the original source was low quality, soft, to begin with (e.g. maybe it was an upscale to begin with) - ie. not really effectively 1080 lines (or close to) of real resolution; or b) the downscale step was a small amount, maybe 1-10% . c) the distribution bitrate was very high. The higher, the more difficult to tell differneces . And if a,b or c are correct and you cannot tell with your eyes, why do you care ?

If you have the known original you can use objective metrics, such as PSNR . This provides irrefutable proof.

Afaik, bit rate values isnt an ideal yardstick in this instance as Ive seen cases where BR has been disproportionately blown up during the transcoding process. PSNR I guess is the only way if you are dead serious about finding out, but that involves stripping down the numbers almost to the point of singularity to obtain the values.

Since there seems to be no easy way to establish this, I think I will have to leave it at that. But thank you to all no less for all your invaluable inputs.