ok, i'll try that. thanks
+ Reply to Thread
Results 31 to 60 of 68
Nice guide. Wish it would work for me. I tried it out on TVLand’s alpha-blended logo. I can get it to erase the logo on ‘frame2’ but it does not work on the rest of the video clip. On frames with a light colored background, the logo is unaffected. On darker backgrounds, the logo tends to ‘fade’ some. The darker the background, the more it fades but it never entirely disappears, even on a black background.
Am I doing something wrong or is it the nature of the video clip I am using? The clip was first recorded in XP mode on a dvd recorder, then imported into VirtualDubMod and saved as an AVI file using PICVideo MJPEG set to 19 and then re-imported back into VirtualDubMod before trying Delogo. I have experimented quite a bit using different settings, different masks, different frames, off-white colors for ‘frame1’, etc but doesn’t help.
Try making up several masks to use. Also, you may need to adjust the white to more of an off white. That's worked for me on tough logos.
Gordian Knot uses virtualdub and avisynth scripts. Can't I do it through Gordian Knot ? I can edit the avs script in there
Here's a script from a few post back
I'm just unsure bout how to put it together
Originally Posted by spanky123
First get and install the delogo filter in Vdub.
The address wasn't given, it's at http://neuron2.net/delogo132/delogo.html
Then follow the instructions in the first post in the thread using VDUB.
In the delogo screen, save the four mask bitmaps in the same directory as your AVI file, using the names "deblend.bmp", "alpha.bmp", "color.bmp", "repair.bmp".
Then take note of the numbers on the sliders of the "repair parameters" to plug into your script.
eg, my Avisynth script for MPEG encoding was:
delogo(0, "", "deblend.bmp", "alpha.bmp", "color.bmp", "repair.bmp", 15, 40, 0, 0)
The number "15" corresponds to a "repair depth" of 1.5, the "40" to "repair power" of 4.0 as was generated by the analysis.
Thanks heaps for that.
I'll give it a shot soon and see how it goes.
good filter. thanks alot.
i can't make it work
in order to enable the filter, the video tab in virtualdub's main vindow must be in "full processing mode" option.right?
well i can load delogo and proceed and see in preview of it that the logo is blurred
but when i hit ok and want to save avi and choose "direct stream copy" in video drop bar, the result is the same as input video and if i choose "full processing mode" video size gets realy huge
so what am doin wrong?
Originally Posted by PUBLICENEMY
To see the results of filters, you need "full processing".
And then you probably want to compress it, so the button below "full processing mode" in the "save as" dialogue click "change" and choose your output format (eg, DivX 5) and the settings for that govern the quality and file size.
sorry for askin again
but theres only 1 option for savin which is "save as avi". on the other hand in video drop bar theres a botton "compression" which u can choose between compression formats or stay uncompressed,
my file is xvid but when i choose xvid it gives me "stats not found" error
what should i do?
also whats the average time for you to complete delogo-ing
LOL, I am stuck on the first step, can someone tell me how to get the all white background the same size as the video (tut is not very specific).l
Do I just make a white .bmp image in photshop?
I dont get it... thanks[size=9]Zhare Your Info Article Publisher
R5DVD.com Support Team
VLCPlayer.info - My Favourite Media Player
Originally Posted by incrediblej
Simplest way: make a screencap (you need several anyway), copy it, fill with white.
I've been researching how to remove (de-logo) watermarks in my videos. I've D/L'ed other
peoples images with watermarks to use as examples -- there's so many types of logos on
stations these days. Anyway.
I'm building a tool to remove logos. I"ve been interested in it for quite some time, prob since
I first came across this discussion back in Sept/2005 but had many other things going on.
The one thing that has stopped me many of times in this endeavor was the fact that there
doesn't seem to be any tutorials on the machanics of logo removal. I mean, I searched for
all kinds of keywords and phrases and still can't come up with any to this day. So, I've
decided to just attempt at it from scratch. Ok, even though this topic does not cover the
mechanics (algos) of how to remove, this discussion is the closest I've come to it and am
just mainly trial 'n error 'ing it as I go.
So far, they way I've gathered it, removing watermarks requires the use of multiple types of
masks, and then some kind of analysis is performed with these masks and then an end result
image without the watermark logo is gone -- or, almost gone. I am hoping to uncover the
mechanical secrets and ultimately learn how to apply it to my tool, and who knows, a plugin
could be next.
If I get closer to the goal I'll post a screen shot of the tool's progress. So until then..
@vhelp - You might start by looking at the source code for the DeLogo filter this thread is about. It's available from the same site as the filter itself, (posted above).
For anyone interested...
I use the following AviSynth script to build my filters files.
A = DirectShowSource("VideoWithLogo.avi", fps = 29.97) B= BlankClip(A, length=30, color=$FFFFFF, audio_rate = 48000, stereo = true).Subtitle("WHITE") B= B++ BlankClip(A, length=30, color=$EEEEEE, audio_rate = 48000, stereo = true).Subtitle("GREY0") B= B++ BlankClip(A, length=30, color=$DDDDDD, audio_rate = 48000, stereo = true).Subtitle("GREY1") B= B++ BlankClip(A, length=30, color=$CCCCCC, audio_rate = 48000, stereo = true).Subtitle("GREY2") B= B++ BlankClip(A, length=30, color=$BBBBBB, audio_rate = 48000, stereo = true).Subtitle("GREY3") B= B++ BlankClip(A, length=30, color=$AAAAAA, audio_rate = 48000, stereo = true).Subtitle("GREY4") B= B++ BlankClip(A, length=30, color=$999999, audio_rate = 48000, stereo = true).Subtitle("GREY5") B= B++ BlankClip(A, length=30, color=$888888, audio_rate = 48000, stereo = true).Subtitle("GREY6") B= B++ BlankClip(A, length=30, color=$777777, audio_rate = 48000, stereo = true).Subtitle("GREY7") B= B++ BlankClip(A, length=30, color=$666666, audio_rate = 48000, stereo = true).Subtitle("GREY8") B= B++ BlankClip(A, length=30, color=$555555, audio_rate = 48000, stereo = true).Subtitle("GREY9") B= B++ BlankClip(A, length=30, color=$444444, audio_rate = 48000, stereo = true).Subtitle("GREYA") B= B++ BlankClip(A, length=30, color=$333333, audio_rate = 48000, stereo = true).Subtitle("GREYB") B= B++ BlankClip(A, length=30, color=$222222, audio_rate = 48000, stereo = true).Subtitle("GREYC") B= B++ BlankClip(A, length=30, color=$111111, audio_rate = 48000, stereo = true).Subtitle("GREYD") A ++ B
EDIT: 05.26.08 Mon AM
Unfortunately, the source is c/c++ and I have no strength to contend with that language
Thanks for trying to help. I'll stick with pascal and wing it on my own from scratch. I know I'm
missing something in my project and that is part of my problem in finding how its done.
EDIT: 05.26.08 Mon 12:56pm
Rather than continue in this one, I might start a new topic on the subject on the development of a
potential watermark removal tool because its more specific about designing the tool and the functions
that make up the process. This way, I can post my progress as I go. Plus, I think I mgiht be on to
WOW! This method of Matt's ROCKS!!! I used delogo back in '03 for awhile and spent (wasted) gobs of time picking majority sections of the show to scan to do the analysis. Now with this "2 frame" method, all that trouble goes away! Thank you, Matt!
But I have one problem. Playing with the "deblend shift" and "deblend falloff" sliders is impossible as my monitor cannot be made bright enough to see the quasi-removed logo at ALL. I really need to adjust the sliders WHILE I have brighter and more normal video behind it. But if I load my video file to do that, the sliders freeze up at their last locations. Of course you did say "Make sure you do this now" before closing the filter for the 1st time. Is there anyway to get around this, short of either creating a new file with the 2 frames at the beginning, or asking the author to fix this?
Another unrelated thought: In '03 I did a great deal of experiementing, making adjustments to all the different features of that big window. I concluded that in all cases, it was best to leave just about everything at default values. This included (I believe) "alpha to repair". While it's true we don't want to blend much of anything, wouldn't it be appropriate to slightly blend away those few cases of some stray sharp pixels (or groups of pixels) that stand out after the filter has been applied? I'm not suggesting we move it toward "repair" but merely leave it on the default of 160. Would be glad to hear any thoughts on this. Of course, I'd need my first problem solved in order to experiment with this 2nd problem.
Use the AviSynth script I posted above and feed it into Vdub, then open the DeLogo filter and search for a "logo on black" frame and sample, then to the end for an all white frame or one of the different gray frames (sometimes gray works better than white) and sample, then you should be able to move to any frame in the video and still make adjustments on the sliders. However, keep in mind that adjusting for one frame might make it worse on a different frame.
Thx for the script. I've tried it but can't figure out how to restart the test for each grey-color without deleting and reloading the filter. I thought that "Reset stats" would do it, but it doesn't. I understand that for each restart of the test (to find which grey level works best), I must sample the grey frame, then sample the frame in my program that has all black surrounding the logo.
And could you tell us which greys have ever worked better for you than the pure white?
BTW, I'm not loading from an AVI; I'm using an MPEG2 file with the plugin MPEG2.vdplugin. I see that by loading it thru Avisynth with Directshow, I lose the ability to jump around via the I-frames, but for this testing, that's okay. Also, Directshow doesn't seem to keep the same frame numbers as Vdub, so going to the same framenumber I had picked in Vdub didn't work.
The output from Avisynth is uncompressed so all frames are key frames. Scrubbing (especially backwards) is often a problem when using directshow to open mpeg2. I use DGIndex for mpeg2 files, works much better (accurate), although it's slower than opening the mpeg2 in Vdub.
I had a clip a while back that Grey1 worked better than White, but I don't remember what station logo it was. White usually works best, but if it doesn't try some of the greys. I think it depends on whether the original logo was done in white or grey, but depending on the level of transparency, it can be really hard to tell how it originated.
And yes, It's also been my experience that I had to delete the filter and start again to try a different grey.
Ive just started experimenting with Delogo and Virtualdub .Im trying to remove a fairly large timeclock on black background at the top of the screen.Im getting good results but not perfect.Any one offer any advice.I was thinking of maybe running the software twice to remove it completely
If it's a timeclock and the time displayed keeps changing what's on the screen this won't work well for you. Anything about the logo that's changing will effect the masks. You would have to make up a mask for each different digit that's displayed and use THAT mask for it only.
Its work DeLogo filter , but i have more simple problems. I use the VirtualDub-mpeg2 and load my dvd file directly. I remove the logo and when i try to run the video the sound is gone. If i try to save as AVI i get bad picture with freezeng frames and no sound.
Its seem to that the video only is loaded without sound.
Do i need some other software to load the video with sound in virtualDub-mpeg2 or ????
It's been a year since I joined VideoHelp in order to interact in this thread. I didn't do much with Delogo all this time, until recently. I am HOT on the trail of something I'm very hesitant to divulge because I don't want to get people's hopes up until I can "deliver".
I'm not a frequent visitor to this board, but I have to thank "Matt D" for the great help his initial post has been to me and to my current project!
I've had an unexpected opportunity to learn some of each of the .NET (dot net) languages, C#, VB, and C++. Took 2 courses in each except C++. (Am presently looking for a job to become a computer programmer once again.) While the Vdub source code (as well as code for most filters) is woefully lacking comments, I am tackling making some significant improvements to Delogo! Not with the essential logic, which is totally beyond my comprehension, but with the tedious mechanics; load this, save this, press button for this, etc. As I speed up the manual config. process, I am getting closer to my goal.
But I'm running into some problems, so I'd like to ask some questions.
1. Has anyone spent any time studying the source code (1.3.2)? BTW, don't bother trying to get help from the author. I had pleasant contact with him via email in 2003. I reached him again recently, asking only that he convert it into VC++ in Visual Studio. (I'm terribly green in C++ and couldnt' do it.) He replied that he doesn't even have a C compiler on his PC now! So we are all on our own. (With the kind help of a VS2005 expert on another forum, I DO have the filter compilable in my VS2008.)
2. Does anyone know WHY the author built his complex system of enabling and disabling features depending on the context?? This was always aggravating to me, but I think I came to see it may have been necessary. Problem is I HAVE to defeat that in order to accomplish some things.
3. One thing I need to do is figure out how to break it enough to allow me to load another video file and still make mask adjustments (Mask contruction parameters). Haven't looked at it too closely, but I would think the only obstacle is that the Analyse mask isn't retained. I believe I can fix THAT. Tinkering with this may required the user to save the 4 masks again, but that's okay. Anyone have any thoughts on this? Possible? Impossible?
4. (A bit unrelated...) I've discovered an important missing step in creating proper masks. I saw one tiny reference to this, but it's not anywhere else. And Matt D's otherwise great tutorial propagated this error. When you create an Analyse mask, you MUST color BLUE most of the image. You leave a reasonable margin of black around the red, but the rest must be blue. In my tests weeks ago, involving "sample frame" from different parts of a 30-sec clip, I was totally mystified as to why a frame delogo'd by my applied mask (based on the 2 BMP trick of Matt D) would get LIGHTER in different places, so that it took on a greyish hue. I was extemely baffled and exasperated! I finally noticed the pattern... My analysis was done as per Matt's plan; sample one frame of pure white, then one where the logo is against a totally black background. But the black frame had a very dark scene. When I switched to a scene with black logo background and MUCH BRIGHTNESS in the REST of the frame, that brightness affects the analysis! Upon discovering this pattern, I figured that's what the BLUE is for. So I broke out the blue paint bucket and covered all but the logo and a small margin around it. Problem SOLVED!!! I think our accidental omissions of this step have been forgiving because (1) Analyses made by scanning the video result in an averaging effect. Not so when you use only 2 frames. (2) Using the 2-frame trick will still work as long as all frames are close to the overall brightness of the non-white frame used in the 2-frame trick. Bottom line... Always paint most of the frame blue!
I may have more questions later. Hope to have good news, too. This tackling of deciphering a Vdub filter has been excrutiating so far, but I'm seeing enough progress to not dislike C++ anymore. (Will still shoot for a job in C#, though.)
I've discovered an important missing step in creating proper masks.
5. Open the image in your preferred image editor (e.g. Windows Paint) and mark the logo and some area around it by a fine red (255, 0, 0) color. Leave some area around the logo as is and mark the remaining area by full blue (0, 0, 255) color.
The red area is the area where the analyser will look for the logo. The logo itself should be a few pixels smaller than the marked area. The blue area will be omitted by the analyser, i.e. the analysis results from the blue area will not be used in any way. The remaining area (original color) will be used as a reference area - the analysis results for pixels in this area will be compared with the analysis results for pixels in the red area to find the alpha and color values. It is therefore a good idea to leave the reference area near the analysed area but not to make it too big.
Vdub related until '07. Matt's helpful thread here DID give examples with no blue, which reinforced my own error even though the docs DO say it, albeit in a way that's not real clear to someone who doesn't fully understand everything stated there. It could have been more explicitly stated as an important step; "Paint everything else blue" and "It's important that you do this". Just stating what's happening (the 2nd part of your quote) is not enough as it makes inferences not automatically in our minds. In fact, you could read it one way to believe that one ought NOT to paint it blue. Well, Karel is from the old Czech republic so the fact it's in english at all is an accomplishment.
ANIMATED LOGO REMOVAL!!!
I have it mostly working! But I don't want to process it against my first movie until I can work out a difficult problem.
I'm stuck on my question #3, the need to do further mask adjustment AFTER I have closed the Delogo window, loaded a new video, and reopened the Delogo window. When you close it, lots of data structures are intentionally erased. I would need to know either:
1. What to do to recreate the "live" analysis data after reloading the filter (and 4 or 5 masks);
2. In lieu of #1, what data to save to a file and reload at the right time.
A few comments about the overall project:
-- As others have discerned, you have to have a separate Delogo instance (and related 4 masks) for EACH uniquely positioned animated logo frame.
-- I have greatly speeded up the manual labor involved in seeking frames and saving masks from within the Delogo configuration window.
-- My test case (a movie I am really eager to finish processing) has a logo appearing 5 times during the movie, with each instance occuring for 877 frames (about 29 seconds). I have 92 instances of Delogo. Instances #1-18 each process only 1 frame. #19 processes frames 18-37 (a stationary portion). #20-69 each do 1 frame (38 thru 87 respectively). #70 processes frames 88 thru 854 (the biggest stationary portion). #71-92 each do 1 frame (855 thru 876 respectively).
-- I have a C# program that greatly facilitates generating the .vcf files that will be used to tweak and process everything.
-- Using this process will still be tedious, BUT I believe I am bringing the work needed down at least an order of magnitude, hopefully to a level that is feasible to actually do.
Perhaps some readers of this thread have not believed I had anything signficant. Whether it is to others, it's significant to me, as it will improve my OWN video collection I am making.
Concerning my question #2, I think I'm backing off on my earlier intent to "break" his system of safeguards. I am trying to learn enough of it to safely working around it and let it stay in place. This means I'm trying to create the conditions that are right for allowing his checks to still operate, yet still allow me to do what I need.
So if anyone has any thoughts about any of this (here or in the earlier posts), please voice them. I envision somehow making Delogo 1.3.3 (or 1.3.3TC) available eventually to everyone (maybe a link in this thread).
Oh, I must warn you (if you haven't discovered this already). Delogo source, while it has a few comments helpful to someone who knows the concepts well, it is largely uncommented and is a nightmare to decode (especially for a C neophyte). But he (the author) is not at fault; he's probably just following the example of the VirtualDub author who has commented NOTHING throughout his whole program. And as the two (Vdub and any filter) are intertwined in the strangest ways, it becomes virtually impossible for anyone who's not a seasoned C++ programmer to confidently and comprehensively figure out what's going on. I've been able to use my long-practiced programming skills in other languages to aide me. But I'm stepping on potentially thin ice with nearly every change I make, thanks to no comments telling what each function does.
Today I found a comment indicating that SourceForge has recently acquired a group called "Ohloh". I found their comment about VirtualDubMod insightful:
"Few source code comments. VirtualDubMod is written mostly in C++. Across all C++ projects on Ohloh, 22% of all source code lines are comments. For VirtualDubMod, this figure is only 13%. This lack of comments puts VirtualDubMod among the lowest one-third of all C++ projects on Ohloh. A high number of comments might indicate that the code is well-documented and organized, and could be a sign of a helpful and disciplined development team."
Having said that, I concede that both Karel (Delogo) and Avery (VirtualDub) have done us all a great service by writing and sharing their work. But it sure is a nightmare to decipher a deeply complex object-oriented application like VirtualDub, when it's so grossly under-commented.
Last edited by TCmullet; 25th Feb 2010 at 16:37.