Scene changes not detected correctly when encoding in x264 vfw codec

[cesm23]

well because 348x288's quality is more than enought to me, and above that the encodings take a longer time, and you know, i am looking for acceptable quality in the videos, not ultra hight and best possible quality... 384 is more than enought especially because the things i capture have a lot of episodes and since i have to encode about 3 or 4 daily i don't want the encoding of each show taking several hours (which usually happens when encoding higher than 384).

And ALWAYS encode from the best source. A 1500Kbps ABR encode as a source is quite bad.
Yes NOW i know that (yesterday in fact) the problem is that i ALREADY have a lot of video in which the best source i have is even worse, since i did a lot of encodings in x264 to abr 1000kbps, i only did 1500kbpps in those at 720x576.... but of course now i know that i won't be using ABR anymore... lo

Oh by the way setting the scene cut threshold at 100 really improved a lot the scene detection in both the vfw version and in the non-vfw version, altough still some are missing but oh well better than going back to xvid :)l

[nm]

well because 348x288's quality is more than enought to me, and above that the encodings take a longer time, and you know, i am looking for acceptable quality in the videos, not ultra hight and best possible quality... 384 is more than enought especially because the things i capture have a lot of episodes and since i have to encode about 3 or 4 daily i don't want the encoding of each show taking several hours (which usually happens when encoding higher than 384).
You don't need a fancy deinterlacer when turning 576i25 to 288p25. Just separate the fields, keep one and discard the other. This gives you 720x288 at 25 fps. Then resize horizontally to 384x288: SeparateFields()
Selecteven()
LanczosResize(384,288)

[cesm23]

Nope, that made the image a lot worse... for example, diagonal stuff gained a lot of "stair" like effect... Want some examples? And yes i disabled yadif in this example :)

[nm]

Yes, this results in a bit more aliasing, depending on how sharp the source actually is.

[cesm23]

Hum... i see... then, since i got much better results with yadif, i can't use this method at least for these source videos... and yes i used the 720x576 mpeg2 as source. Anyway i am pretty satisfied with the results with yadif i really hope i don't find any part of the videos incorrectly deinterlaced :)

[Dark Shikari]

For example i found out that by using a CRF value of 19 the video i converted from the mpeg2 source has 580mb. When i encoded the same video (the source) at ABR 1000kbps the size is 611mb and i could notice that in the abr version it had a very litle less blocks in the image than the crf at value 19. Any tips of what more i should compare more exactly ?If you're managing to get blocks at CRF19, you're seriously doing something wrong, as you should need a magnifying glass to tell the difference between CRF19 and the source.

Can you post comparisons of the same frame between ABR, CRF, and the source?

[cesm23]

oops you were fast replying the message :)

No wait, before the screenshots... regarding the source, i send a screenshot of the video using the avisynth in avsp (and using the save image function) script with the yadif and 384x288 resize ? Or i should take that screenshot from the mpeg2 file directly, still being interlaced and in 720x576?

By the way yes i needed to amplify the video to notice the blocks i ONLY did this because i found it exactly the same as the abr version that's the only reason i complained, because i don't mind being like this, i only find it strange that both videos seem to have the same quality and only by amplifying the video with a glass i could notice some differences.

Besides the mpeg2 source already has some blocks of it's own, my point here is only comparing the crf and abr methods.

[Dark Shikari]

By the way yes i needed to amplify the video to notice the blocks i ONLY did this because i found it exactly the same as the abr version that's the only reason i complained, because i don't mind being like this, i only find it strange that both videos seem to have the same quality and only by amplifying the video with a glass i could notice some differences.When you've saturated the bitrate to the point where you cannot tell the difference anymore, you won't be able to tell the difference between settings at all.

If that's the case, you're wasting bits. Raise the CRF.

[cesm23]

Oh ok, thanks :) I will keep that rule in mind now when trying to find the correct value. And like this i can have a smaller video without noticing much diference, which is what i want.

Jeez i can't believe that FINALLY i found the most adequate settings to encode in x264 for my specific video captures quality, thanks to all these replys, i thank you and the others here for their support, and forgive me if i did too many questions... I hope i don't have more problems that i can't solve by myself for now. And at least the scene change detection problem is now partially solved now that x264 with the threshold at 100 it detects almost all scene changes i need, it misses only a few now.

[cesm23]

Ok now i finally did some encodings of full tv shows in CRF 19 quality, and needless to say i am satisfiyed with the results, i am just having a small doubt... since i still don't know exactly how crf is supposed to work, and even tough i read it's description in the x264 settings guide, i am just a bit afraid : for example using the exact same settings i recorded two different tvshows in 384x288 in crf 19, i noticed that one has a average bitrate of 730 and the other 1600kbps... Is it normal such a big difference ? I shouldn't worry about the one that has 730kbps ? I am just bit worryed about the one in 730kbps even tough i am now encoding it at abr 1000 too just to see if there is any higher quality loss (you see all this time before i used CRF i always guided myself with the visual quality of the abr1000 encodings) than when i did the original crf 19 vs abr1000 comparings, but i hope not, since i don't want to be testing new cfr rates for each video individually!

But anyway this is normal right ? This happened just because the one that had a higher bitrate has more motion scenes and therefore it needed higher bitrate right? I just hope there isn't much quality loss in the one that got a 730kbps average rate.
Because if the visual quality is the same in both videos and it just increased bitrate in the other one with more bitrate to keep the same visual quality, and if the one that got lower bitrate was able to maintain the same visual quality even at a lower bitrate, then indeed this crf mode is GREAT!! :)