[jmarshall]

just a numbers update:

so far i have received input from 2 other people that have taken part in the test.

please take part if you can - this is an important part of xbmc and we want to improve the quality as much as we possibly can. in order to do this, however, we must have a set of decent clips and methods to test with, and a decent number of people willing to take part in tests.

if you can't see any difference in the test file i posted, then please let me know - this is just as important as if you do see a difference!

also, if you have other files where you feel you can see differences more clearly, then please make them available so that i can take a look at them.

thanks,
jonathan

[jmarshall]

ok, i've received 10 entries now, which i think is almost enough to get some decent data out of it, as there are some trends occuring.

i've decided to leave it until sunday night to reveal the results, giving folk who will have more time over the weekend to have a shot. i'll reveal on sunday at midnite gmt (or thereabouts).

cheers,
jonathan

[poing]

can hardly wait for the results!

[jmarshall]

thanks for all of you that took time to take part in the test.

the results are as follows:



obvious diff slight diff no diff preferred bottom preferred top top method bottom method
1a 4 3 8 2 4 hqps hwo
1b 4 2 8 6 hwo hwo
1c 5 4 6 8 hwo hqps

2a 8 2 5 5 5 hwo hqps
2b 5 3 7 2 5 hqps hwo
2c 2 2 11 2 1 hqps hqps

3a 3 1 11 3 hwo hqps
3b 4 2 9 1 4 hwo hwo
3c 3 3 9 1 4 hqps hqps

comments:

clip 2a seemed to be the one that most people found a difference. opinions were evenly split over which one was preferred, however. interesting that clip 2b which was identical with the clips reversed did not show the same amount of perceived differences.

clips 3a-3c showed very little differences (even in clip 3a where they were in fact different) possibly due to the amount of movement in the images and the colour variation.

the fact that there was almost unianimous opinions that the top one was better is interesting, given that the clips were uniformly spread between top and bottom (possibly a product of the screens you were viewing them on?)

in the preference stakes, it was a dead heat, with those that found differences liking the hwps 17 times, and the hwo 17 times.

overall, i don't think anything can be honestly drawn from the results - many of you found differences when do differences existed (ie same clip top and bottom) and many found no differences where there was a difference. perhaps the clip wasn't of the desired type to show a difference, or perhaps the fact that they were stacked vertically made it harder (on a laptop for instance, the backlighting makes way more of a difference to colour than the quality of the image) - i'm interested in hearing your opinions on how to better improve the test.

those that took part may now wish to comment on their success (or otherwise).

cheers,
jonathan

[GhostDog__]

interesting results. i had a feeling that the same render method was used for both the top and bottom video in some of the segments. i thought that perhaps 2a used the same renderer for the both top and bottom vids.

would it matter in anyway what video player, codecs each tester used? i used media player classic with the k-lite mega codec pack (v2.68). it uses ffdshow [build 2005-12-22] for playback of xvid and divx files. i'm wondering if perhaps each tester should be using the same player, and codecs.

i kind of wish i had a crt monitor instead of a lcd just for this test.

[simba115]

i was under the impression my xvids looked quite poor on my xbox so i've just now stumbled upon this discussion that's been going on for an incredible amount of pages.  i tried my best to go through all 30+ pages of this thread plus more pages from other threads to try and figure out how to get my xvids to look like they do on my pc and a lot of it was wayyy too technical for me. i'm hoping someone can paint a clear picture of what the deal is.

-can we get an xvid movie to play at the same quality on the xbox as a pc connected to a tv? or because of hardware/software or other limitations of xbox this is not possible?

-am i correct to say that the settings to get the best picture so far should be to choose hardware overlay renderer, no post-processing filters and no softening. (i've already switched from high quality pixel to hardware overlay and there is a much better picture)
*if these settings are not correct, can someone post their 'best picture' settings for xvids?

-are these above settings enough to get the picture to be at least as good as on the old xbmp? (i am a new xbox owner, i have always lived in a world with xbmc) but i read numerous posts about the better picture quality with the 'old' 2.4 player.
-this old xbmp that you speak of, can't i just copy it over to the f/apps and use it to play the movies instead of the xbmc and problem solved? (or will i totally screw up my dash trying to do this?)
-does it make a difference if the xbox is connected to the tv thru component or composite ( i couldn't see one ) when playing xvids?

thanks for this

[Gamester17]

can we get an xvid movie to play at the same quality on the xbox as a pc connected to a tv?

please understand and respect that this topic thread is only about the renderer issue(s); so if switching renderer from pixel-shader to hardware-overlay did help you (when comparing xbmp vs. xbmc with post-processing filters disabled in both) then i don't think that the renderer is your problem thus you are off-topic here (which you would have known if you took the time to read every post in this whole thread).

...my guess is that your real problem is that your xvid's use qpel as the mplayer-core have issues with qpel in xvid, please read this thread instead (and reply to it if need), tia

am i correct to say that the settings to get the best picture so far should be to choose hardware overlay renderer, no post-processing filters and no softening.

add disabled the flicker-filter (ie. setting it to "0") under video settings then yes you are fully correct, that is all you can do in xbmc to get the best quality (that i know of).

ps! please respect the fact that we don't support xbmp anymore so please don't ask questions on how to use it here, if you need help on xbmp then go to third-party forums like on xbox-scene.com

[Dolfhin]

so, the difference between overlay and pixel shader is as good as non-existing and xbmc is for most people on par now with xbmp?

[frov]

to react on dolfhin:

well, i was one of the early 'complainers' on xbmp being better compared to xbmc. back in those days.

then the good work was done by jm some months ago with the introduction of this 'hardware overlays' and 'hq pixel shader' and such. for me, back then comparing one on one, the fight was over: xbmc was at least as good as and sometimes even better (sharper) than xbmp. exit xbmp for fog-inhabited movies (like master and commander).

i also joined the test last weekend and gave my results. basically i indicated that i did not see significant differences whatsoever. so whoever is complaining nowadays on xbmc quality, i do not see it with my setup (being a normal good tv with s-video from the xbox).

so i am pretty happy as it is.

[simba115]

...my guess is that your real problem is that your xvid's use qpel as the mplayer-core have issues with qpel in xvid, please read this thread instead (and reply to it if need), tia

add disabled the flicker-filter (ie. setting it to "0") under video settings then yes you are fully correct, that is all you can do in xbmc to get the best quality (that i know of).

ps! please respect the fact that we don't support xbmp anymore so please don't ask questions on how to use it here, if you need help on xbmp then go to third-party forums like on xbox-scene.com

sorry. after reading most of the posts here i thought i was on topic    but thanks for your help. i'll try the other thread for more info.

[GhostDog__]

so i am pretty happy as it is.

so am i. your absolutely right about xbmc's video quality being on par or even better than xbmp's video quality.

xbmc's video quality has come a long way. i'll be completely satisfied even if nothing further can be done to improve the video quality.

i just commend the developers for looking into ways of making the video quality even better if possible. i'm happy to help in any way i can.

[Markeen]

i took part in jmarshall test and i must say i was unable to make any quality jugement. all the videos looked pretty much the same to me in terms of quality.
but last night i was toying around the quicktime trailer script and found a video where i could tell the difference between the hardware overlays and the hq pixels shaders.
look at this quicktime trailer ('when a stranger calls' large) right after the pg warning and before the phone appears there is a black screen. with hqps it looks plain black but with the ho this black screen is filled with almost black rectangles. it maybe hard to seen but it is there.

can anyone else see that ? i am using yesterdays build (02-fev-06).

[carlosrealtor]

the quality drastically drops on certain dark scenes. the dark areas of the scene, often dark backgrounds, have the colour range drop away to almost flat colour blocks and colour banding and colour stairstepping is present. detail is also lost.the light scenes are excellent, no issues at all, its just these dark scenes, where im guessing the quants are of really high values

contrastingly to others posts i noticed that overlay renderer showed very little difference to the other renderers quality. the problem is present in all. smoothing made no noticable difference except obviously taking the sharpness out. the brightness, contrast, alpha sliders couldnt be tweaked to remove the issue.

also forgot to mention that if the same scene is played back on a pc there are none of these issues at all.

hello, short-time reader first-time poster here. i've been using xbmc for a while here to watch dvd's and xvid's and i've been noticing the exact same picture quality issues the gentlemen whom i quoted below so accurately described. so i wanted to echo his comments in the hopes that there is a solution out there to this seemingly specific problem.

at first i thought i was just seeing compression artifacts that weren't apparent on my monitor, but now it seems fairly obvious to me that this is an issue with either the xbmc's media player or its xvid codec or the interaction between the two. i have a very low level understanding of video encoding and decoding, but it looks to me as if the player is not able to discern between the different shades of gray near black. so when it decodes the image, you end up with a visible border between black and the next brightest shade of gray. and you see color banding, similar to this:

http://en.wikipedia.org/wiki/color_banding

is there any reason why this would occur with the xbmc's media player but not with media player classic, which i use on my computer?

[Gamester17]

@carlosrealtor, please read my previous post (posted jan. 24 2006,13:05) that was addressed to simba115; xvid is not reliable for testing here, (nor are those xvid issues related to any potential renderer issues). instead test with dvd-video and/or vob-files from a dvd-video disc, if you can replicate your issue with those then please cut a short piece of video (that the problem can be clearly seen with) and submit it to us (see faq), logs and setup-information (your hardware/software settings) is also needed.

[alexh]

something has changed with the high quality pixel shader since i initially came up with a sample (start of pirates of the caribbean xvid - vite) which dramatically showed the differences.

hqps now shows non of the artifacts it used to. something has changed. not sure which version, but as of 2006-05-28 it has definately improved.

it would be interesting to know what was done.