Quantcast Hcmatic: Scene Change Detection - digitalFAQ.com Forums [Archives]
  #1  
05-04-2007, 09:44 PM
supermule supermule is offline
Free Member
 
Join Date: Sep 2005
Location: Donkeyland
Posts: 210
Thanks: 0
Thanked 0 Times in 0 Posts
Why is the HCmatic not using the scene change detection in its prediction routine and then the encodes ???

KVCD specifications suggests usage of scene change detection, if the encoder permits.
Reply With Quote
Someday, 12:01 PM
admin's Avatar
Site Staff / Ad Manager
 
Join Date: Dec 2002
Posts: 42
Thanks: ∞
Thanked 42 Times in 42 Posts
  #2  
05-05-2007, 03:40 AM
Dialhot Dialhot is offline
Free Member
 
Join Date: May 2003
Posts: 10,463
Thanks: 0
Thanked 0 Times in 0 Posts
Because SCD completely screw the prediction and I always said that it's nut to use it. It was the case with CQMatic, and it's still the case.

Prediction is based on slices (in case of HCMatic : 1 second each minute of movie). At each slice end there is a very big chance that the SCD algorythm detects a scene change and insert an I frame. The result is that the % of I frames in the encoded sample is much more important than the reality in the whole movie. Thus the sample size is bigger that it should be, and in the end the Q value found is way under the ideal value.
Reply With Quote
  #3  
05-05-2007, 06:24 AM
rds_correia rds_correia is offline
Free Member
 
Join Date: Apr 2003
Location: Chinese Democracy starts now!
Posts: 2,563
Thanks: 1
Thanked 0 Times in 0 Posts
Quote:
Originally Posted by Dialhot
Because SCD completely screw the prediction and I always said that it's nut to use it. It was the case with CQMatic, and it's still the case.
Exactly!
It doesn't matter if you're using HCMatic, CQMatic or even a manual prediction approach.
If you are running a prediction routine you have to disable the scene change detection otherwise you'll get mistaken output figures for Q.
As for enabling SCD for final encoding I just don't know...
Sometimes I do that and I get a bit oversized movies that I have to run through DVDShrink or then reencode but sometimes I do that and the outcome is pretty close to the final predicted sizes.
So IMHO there's no sense in using SCD at all.
Cheers
__________________
Rui
Reply With Quote
  #4  
05-05-2007, 08:24 AM
supermule supermule is offline
Free Member
 
Join Date: Sep 2005
Location: Donkeyland
Posts: 210
Thanks: 0
Thanked 0 Times in 0 Posts
Quote:
Originally Posted by Dialhot
Because SCD completely screw the prediction and I always said that it's nut to use it. It was the case with CQMatic, and it's still the case.

Prediction is based on slices (in case of HCMatic : 1 second each minute of movie). At each slice end there is a very big chance that the SCD algorythm detects a scene change and insert an I frame. The result is that the % of I frames in the encoded sample is much more important than the reality in the whole movie. Thus the sample size is bigger that it should be, and in the end the Q value found is way under the ideal value.
Thanks for the info!!
Reply With Quote
  #5  
05-11-2007, 12:07 AM
Boulder Boulder is offline
Free Member
 
Join Date: Sep 2002
Location: Lahti, Finland
Posts: 1,652
Thanks: 0
Thanked 0 Times in 0 Posts
Quote:
Originally Posted by rds_correia
As for enabling SCD for final encoding I just don't know...
Sometimes I do that and I get a bit oversized movies that I have to run through DVDShrink or then reencode but sometimes I do that and the outcome is pretty close to the final predicted sizes.
So IMHO there's no sense in using SCD at all.
Cheers
It's strongly recommended that scene change detection is enabled. HC has a very good algorithm which will catch them very well. Otherwise you might end up in situations where a scene change is P-B which means that there will be severe degradation of the image quality before the next I-frame arrives.
Reply With Quote
  #6  
05-11-2007, 02:27 AM
Dialhot Dialhot is offline
Free Member
 
Join Date: May 2003
Posts: 10,463
Thanks: 0
Thanked 0 Times in 0 Posts
Quote:
Originally Posted by Boulder
Otherwise you might end up in situations where a scene change is P-B which means that there will be severe degradation of the image quality before the next I-frame arrives.
You are right. I never think about that. Not sure this is noticiable, at least, I never have been chocked by that ((with my eyes, while watching at the video I mean), but I will think about activating it.
Reply With Quote
  #7  
05-11-2007, 03:09 AM
Boulder Boulder is offline
Free Member
 
Join Date: Sep 2002
Location: Lahti, Finland
Posts: 1,652
Thanks: 0
Thanked 0 Times in 0 Posts
In prediction, it is better to disable it as you already had figured. But in the regular encode, there's really no good reason not to use it..I recently did a test in which I noticed what happens when scd is not used or doesn't work. There was a fast action scene followed by a lo-motion scene, the scene change was P-B as CCE didn't for some reason put an I-frame there. The first few frames of the latter scene looked really terrible as they were predicted from incorrect motion vectors due to the missing I-frame. It was quite noticable during playback, in fact it was why I started looking at the scene more thoroughly.
Reply With Quote
  #8  
05-11-2007, 11:32 PM
rds_correia rds_correia is offline
Free Member
 
Join Date: Apr 2003
Location: Chinese Democracy starts now!
Posts: 2,563
Thanks: 1
Thanked 0 Times in 0 Posts
Thanks for your comments Boulder .
Cheers
__________________
Rui
Reply With Quote
  #9  
05-19-2007, 08:21 AM
Dialhot Dialhot is offline
Free Member
 
Join Date: May 2003
Posts: 10,463
Thanks: 0
Thanked 0 Times in 0 Posts
I'm currently regarding a video that I did not encode where SCD is obviously off, and the result is awfull. Each SC the frame is at best pixellated, at worst awfullyblended.
I will add the SCD to final encode in HCMatic once I verify that does not screw up the final filesize too much.
Reply With Quote
Reply




Similar Threads
Thread Thread Starter Forum Replies Last Post
Avisynth: Dynamic Linear Adaptive Filtering and Scene Change Detection supermule Avisynth Scripting 3 09-15-2006 01:45 AM
Avisynth: Scene Change Detection? Prodater64 Avisynth Scripting 1 02-29-2004 09:05 PM
Detect scene change big Conversion et d'Encodage de Vidéo (Français) 1 12-27-2003 12:48 PM
KVCD: Scene change detection? Boulder Video Encoding and Conversion 1 01-27-2003 11:50 AM
KVCD: Detect Scene Change? Pike Video Encoding and Conversion 1 08-08-2002 12:56 PM

Thread Tools



 
All times are GMT -5. The time now is 11:25 AM  —  vBulletin © Jelsoft Enterprises Ltd