faq.txt

Video IO toolbox for matlab. 用directshow做的

##############################################################################
#### Frequently Asked Questions
##############################################################################

------------------------------------------------------------------------------
Q: Why should I use videoIO?

A: See README.txt's "Motivation" section.

------------------------------------------------------------------------------
Q: How do I use videoIO?

A: See INSTALL.dshow.html (Windows) or INSTALL.ffmpeg.txt (Linux) for 
   installation instructions.  Then look at the documentation in
   the constructors @videoReader/videoReader.m and @videoWriter/videoWriter.m
   by typing
     help videoReader
     help videoWriter
   in Matlab.  A simple videoReader example is given in README.txt and 
   tests/videoWriterDemo.m is a good example of writing videos:
     edit(fullfile(videoIODir, 'tests', 'videoWriterDemo.m'));

------------------------------------------------------------------------------
Q: videoIO, videoReader, and videoWriter: what's what?

A: Originally, this library was just called videoReader because it was only
   used to stream videos into Matlab using mex function plugins to access
   3rd party libraries like ffmpeg and DirectShow.  Later we found it 
   natural to  extend the capabilities to include writing videos as well.
   videoIO is the name of the library that allows reading videos with 
   videoReader objects and their plugins and writing videos with videoWriter
   objects and their plugins.

------------------------------------------------------------------------------
Q: How do I install videoIO?

A: See INSTALL.ffmpeg.txt on GNU/Linux and similar operating systems and
   INSTALL.dshow.html on Microsoft Windows.

------------------------------------------------------------------------------
Q: Why are precompiled binary mex files only supplied for Windows?

A: With respect to core library interfaces, Windows is a very stable and 
   consistent operating system and thus the provided binary mex files are
   likely to work on any Windows XP or Vista system.  
    
   There are vast differences in versions of ffmpeg, gcc, and base system
   libraries between different Linux distributions and different releases of
   a given distribution.  To avoid maintaining a large set of binary 
   packages, we instead provide the source and step-by-step installation
   instructions (see INSTALL.ffmpeg.txt).

------------------------------------------------------------------------------
Q: Why should I install videoIO on a local hard drive and not on a 
   networked location like my home directory when using Linux?
   
A: It is very easy to have different versions of ffmpeg running on different 
   machines, especially if you have compiled from sources and/or if you have
   some 32-bit and some 64-bit machines.  There are incompatibilities between 
   different versions of ffmpeg that can be hard to track down if the binary 
   version of a videoIO plugin was built against a one version of ffmpeg's 
   libraries and loaded with a different one.  These include changes in
   data structures that are not detected at runtime and different sets of
   supported codecs.  Thus we strong recommend *not* placing videoIO in a 
   networked location such as your home directory.  See INSTALL.ffmpeg.txt for
   some suggested locations.
   
------------------------------------------------------------------------------
Q: What is the difference between the ffmpegPopen2 and the ffmpegDirect 
   plugins on Linux?

A: For a comparison of the two plugin sets, type 
     help videoReader
   and/or 
     help videoWriter
   in Matlab.

------------------------------------------------------------------------------
Q: When I build videoIO on Linux, I see warning messages of the form
      Warning: You are using gcc version "3.3.5".  The earliest gcc version 
      supported with mex is "4.0.0".  The latest version tested for use with 
      mex is "4.2.0".  To download a different version of gcc, visit 
      http://gcc.gnu.org 
   
A: If your version of gcc is incompatible with the one used to build Matlab,
   very strange errors can occur when running mex files.  We have created the
   ffmpegPopen2 plugin specifically to address this issue.  Use it instead of
   the ffmpegDirect to be safe.

------------------------------------------------------------------------------
Q: The linker can't find sws_freeContext, sws_getCachedContext, or sws_scale
   on Linux.  What's wrong?

A: Rebuild ffmpeg from the sources using the --enable-swscaler flag when
   running ./configure.  See INSTALL.ffmpeg.txt.

   These are swscaler functions used to do image scaling and color conversion.
   Starting with version 51.11.0 of libavcodec, ffmpeg has deprecated 
   img_convert (a more user-friendly, but less powerful and less efficient 
   colorspace converter) in favor of using swscaler.  Since the public header 
   files give no mechanism for determining whether ffmpeg was compiled with 
   swscaler vs. img_convert support, we assume swscaler is used when 
   libavcodec's version is at least 51.11.0 and that img_convert is to be used
   for earlier versions.  
   
------------------------------------------------------------------------------
Q: I had some error with one videoReader or videoWriter object and now all
   of the rest of my opened videos are having troubles and/or I'm having 
   troubles opening new ones.  What do I do?

A: Best case:   try opening the video file again.
   Medium case: type "clearVideoIO" -- this will unload the plugins
   Worst case:  type "clear all" -- this will unload the plugins, close any 
                  opened/locked files, etc.
   
   Our plugins attempt to detect when a codecs may have done a bad thing 
   like produce rouge threads, segmentation faults, or other crashes.  If we
   think an unrecoverable error has occurred, the library may automatically 
   close all open video files.  Typically, we are able to detect a problem 
   and cleanly unload the affected plugin.  In such cases, videoWriter 
   objects are properly destructed so that valid video files are created.  
   Doing the unloading unfortunately invalidates all open video handles 
   within the same plugin.  Our library tries hard to avoid these situations 
   and avoid unrecoverable errors as much as possible, but they still occur 
   under some situations.

------------------------------------------------------------------------------
Q: The installation instructions tell me to modify my environment variables.
   What are they and how do I do this on Windows?
   
A: "Environment variables are a set of dynamic values that can affect the way 
   running processes will behave on a computer 
   (http://en.wikipedia.org/wiki/Environment_variable)."

   While setting up compiler options, we need to add or modify a few 
   environment variables.  This does things like telling Windows where to 
   find programs that will be executed on the command line and where DLLs 
   are located.

   In Windows Vista, setting environment variables is very similar to the way 
   it's done in Windows XP:
      http://support.microsoft.com/default.aspx?scid=kb;en-us;310519
   Here we describe the general procedure for Windows Vista.

   For the sake of concreteness, we'll run through an example.  Suppose you 
   have installed Visual Studio on a 64-bit operating system to its default 
   location,
      "C:\Program Files (x86)\Microsoft Visual Studio 8"
   We'll want to have command-line access to the compiler utilities, so we 
   want to add the "VC\bin" subdirectory to the system path.

      1) Click on the start menu then right-click on "Computer" and select 
         "Properties"

      2) Click on "Advanced System Settings" in the Tasks list on the left

      3) Go to the "Advanced" tab and click the "Environment Variables" button

      4) Edit the "Path" system variable

      5) Add ";C:\Program Files (x86)\Microsoft Visual Studio 8\VC\bin" 
         (without quotes) to the end of the existing string.  Semicolons (;) 
         delimit different paths, so if you add a new path to the beginning of
         the string, you'll need to put the semicolon at the end of this added
         string instead of at the beginning.

      6) Close all dialogs and windows you've just opened (accepting changes, 
         not canceling them).

   Sometimes we will use command-line syntax to indicate changes to the 
   system environment variables, e.g.
    
     set PATH="%PATH%;C:\Program Files (x86)\Microsoft Visual Studio 8\VC\bin"
      
   should be interpreted as a request to do all of the steps listed above.

   Important note: spaces matter in most environment variables.  For example,
   in the PATH above, it is important not to put any whitespace before or 
   after the semicolon.

------------------------------------------------------------------------------
Q: When setting up environment variables on Windows, can I use the user 
   environment variables instead of the system variables?
   
A: Yes you may.  Here's why we suggest using the system variables...
   We assume that most Windows boxes are either single-user boxes (in which 
   case it doesn't matter at all) or all users will want to use and build the 
   library.  In our experience, it's much easier to set the system variables 
   once than it would be to set up user variables for every single user.

------------------------------------------------------------------------------
Q: How can I tell if I have 32-bit or 64-bit Windows?

A: The easiest (to explain) way is to look and see if you have a folder called 
       C:\Program Files (x86)
   If this directory exists, then you are using 64-bit Windows.  This 
   directory is where 64-bit Windows puts 32-bit programs.

------------------------------------------------------------------------------
Q: Long seeks are not frame-accurate.  For example, on Windows
   
    vr = videoReader('numbers.divx611.avi');
    seek(vr,200); I = getframe(vr); 
    figure(1); image(I); % Shows 199 - inaccurate
    vr = close(vr);

    vr = videoReader('numbers.divx611.avi');
    seek(vr,100); 
    seek(vr,200); I = getframe(vr); 
    figure(2); image(I); % Shows 200 - accurate
    vr = close(vr);

   What is happening here?

A: When making the specification for decoders, Microsoft chose to not require
   frame accuracy.  For variable bitrate codecs, this means that some decoders
   will choose to speed up their seeks by estimating the file location of the
   desired frame and then they just read the first frame they find near the
   estimated location.  For long videos, these imprecise seeks can be far 
   faster than precise ones.

   Currently, you can guarantee frame-accurate seeks by using the next method
   instead of the seek method or by adjusting the 'preciseFrames' constructor
   parameter when creating a videoReader object.  Type "help videoReader"
   at the Matlab prompt for details on setting 'preciseFrames'.

------------------------------------------------------------------------------
Q: I am using the DivX decoder on Windows and all of the frames are off by one
   position.  
   
   For example, if I open a new file and call next() twice, I still get the 
   first frame when I call getframe().
   
   What is happening?

A: By default, the DivX decoder chooses to delay all output by a single frame.
   To modify this behavior, go to the Start menu and pick the DivX folder and 
   the "DivX Decoder" directory under that.  Launch the "Decoder Configuration
   Utility."  In the Decoder tab, uncheck "Reduce CPU spikes using one frame
   delay" and "Use advanced hardware overlay."  Both options must be unchecked
   to remove the problem.
   
   An alternative solution is to use ffdshow-tryouts to decode DivX files on 
   Windows.  As of Nov. 2007, these decoders can be found at
   
     http://ffdshow-tryout.sourceforge.net/
   
   These decoders can decode a wide variety of compression formats and are
   quite configurable.  

------------------------------------------------------------------------------
Q: What combination of 32-bit and 64-bit software can be used with the 
   videoIO toolbox?

A: On 32-bit Windows, only 32-bit Matlab and 32-bit DirectShow with 32-bit 
   codecs are supported.

   On 64-bit Windows, one may use 32-bit Matlab with 32-bit DirectShow and 
   32-bit codecs.  Alternatively, one may use 64-bit Matlab with 64-bit
   DirectShow and 64-bit codecs.

   On 32-bit Linux, only 32-bit Matlab and 32-bit ffmpeg are supported.

   On 64-bit Linux, nearly any combination of 32-bit and 64-bit Matlab and
   ffmpeg may be used with each type of plugin.  See INSTALL.ffmpeg.txt and
   makefile for details.