Hi,

Take a look at %windir%\directx.log. See if you see something in there that
could help you find out the reason.

Here are some other things to try:
1, Are you running early betas of VC7? If so, remove these.
2, Jan 2002 version of MSDN library has some known incompatibilities with DX SDK
setup. Search MS Knowledge base if you want more details. Upgrade to a later
version of the library or remove this version.
3, If you have a previous version of the DirectX SDK, then uninstall it before
installing the new version.

Vasi





                                                    To:   OpenCV@yahoogroups.com
                                                    cc:   (bcc: Vasanth
Philomin/ACN/RESEARCH/PHILIPS)
                                                    Subject:    Re: [OpenCV]
DirectX 9.0b installation problem?

                Alexander Shyrokov                  Classification:
                <shirokov@...
                u>

                07/31/2003 06:15 PM
                Please respond to OpenCV






I did and it changed nothing. I updated all I could... and it is written
that it should work for 2000 professional. And it does not even show
why. Maybe there is a way to see the reason?

vasanth.philomin@... wrote:

> I am not sure if this fixes the problem, but I would try to install
> windows 2000 sp4 first and then retry the DirectX 9.0b installation.

--
Regards,
      Alexander.    http://sjcomp.virtualave.net



Change settings: http://www.yahoogroups.com/mygroups, select
    Get Emails (get all posts)
    Daily Digest (one summary email per day)
    Read on the web (read posts on the web only)Or Unsubscribe by mailing
OpenCV-unsubscribe@yahoogroups.com


Your use of Yahoo! Groups is subject to http://docs.yahoo.com/info/terms/

--- In OpenCV@yahoogroups.com, "linbaolon" <linbaolon@y...> wrote:
> Where can I get IPP 3.0 ?
> Does it free?

No, you will have to buy it from Intel.
it cost about $200?.

The old version Intel Image Processing Library (Intel IPL) was free,
and OpenCV 2.x will work with that.  But you might find it difficult
finding the binary.

OpenCV 3.x should work without IPP? but it will probably run slower?

Hi,
   can any body tell me about calculating mahalanobis distance using
cvMahalanobis function,when I called cvMahalanobis method, it is
returning with an error 'Improper type parameter'.

   here iis the code:-

     cvMahalanobis(Rec_Img_Feature,mean,CoVarMat);
     Rec_Img_Feature is a 1D matrix contains features of an Image,
     mean is also 1D matrix conating the mean values,   and
     CoVarMat is 2D matrix containing sampled features of image.
Error appeared me as Bad type parameters

    So I changed parameters into CvMat type,
with the help of cvmSet method... Now it is showing parameter formats
are not matching ?

I don't know about these format, Could any body help me in this...

  Thank U


sreekanth

I have been fool enough to pay for the IPP,
but none of my old IPL functions or variables work e.g. IplImage.

Is there anyway to use these old types or should I port all my code
to use OpenCV functions/types?

Even then how can I be sure that openCV is using the IPP, or even
knows I have it installed?

Sorry, just to answer my own question!!

OpenCV 3 actually contains "most" IPL functions and structures or
very similiarly named and functioning alternatives.
So if I have a compile error due to an old IPL function/structure
etc. I should make sure things like Cv.h and cverror.h are included
at the top of the file.


I'm just going to check out the "CV_TURN_ON_IPL_COMPATIBILITY"
section as well

cheers



--- In OpenCV@yahoogroups.com, "beefheart36" <testtest36@h...> wrote:
> I have been fool enough to pay for the IPP,
> but none of my old IPL functions or variables work e.g. IplImage.
>
> Is there anyway to use these old types or should I port all my code
> to use OpenCV functions/types?
>
> Even then how can I be sure that openCV is using the IPP, or even
> knows I have it installed?

Hello,

There is sample called "Image Processing Sample", which is IPL rewritten using
IPP.
I was sure it is in IPP 3.0, however after getting your message I hit the Intel
site and found out, it will arrive only in IPP 4.0 (now it is
distributed as a part of beta program:
http://www.intel.com/software/products/ipp/beta/imagesamples.htm ).
Sorry if it was me who made you confused about that, I am really confused too
(as I use IPP rarely and last time
I took some internal build of IPP 3.0 beta).

But you can still get benefits from IPP 3.0 (besides that you could get IPP 4.0
for smaller price).
OpenCV makes use of IPP. Beta 3.1 uses only ippcv (computer vision primitives;
it gives faster pyramids, morphology, template matching ...).
The next update that is on the way, will support other parts of IPP, such as
ippi (image processing, base for IPL),
ipps (signal processing) and ippvm (math functions).

To make OpenCV use IPP, put CPU-specific DLLs (*w7.dll for P4, *a6.dll for PIII,
*m6.dll for PII or old Celerons) to the system
path, e.g. to C:\WINNT\System32, or, which is better, add
<ipp_install_path>\bin\ipp20 to the system path
(MyComputer->Properties->Advanced->Environment Variables). Notice that if you
run code from within Developer Studio, you will
need to add the path to "Executable files" folder list (DevStudio menu
Tools->Options->Directories->Executable files).

To check if OpenCV makes use of IPP, insert the following code in your program
(you will need to include "cv.h" and link against cv.lib):

char* version = 0;
char* loaded_modules = 0;
int loaded = 0;

cvGetLibraryInfo( &version, &loaded, &loaded_modules );

if loaded is not 0 after the call, OpenCV found and loaded some of IPP modules.
The names of the loaded modules should be written in
the loaded_modules string.

Then, recently Jerome Landre posted a nice tutorial, how to use OpenCV and IPP
together. The tutorial is linux-oriented, but all the code
samples should work well on Windows as well. The tutorial (ippocv.pdf) can be
downloaded here:
http://prdownloads.sourceforge.net/opencvlibrary/ippocv.pdf?download

Finally, if you send me the names of IPL functions that are most important to
you (from which you need the maximum speed)
and the data types you are work with, e.g.
"I need iplDCT for IPL_DEPTH_8U input and IPL_DEPTH_32F output" or "I need
iplWarpAffine for IPL_DEPTH_8U, 3-channel images"
I can prepare the small wrappers that take IplImage's and call IPP inside. As
you will see, it is quite easy to do.

About OpenCV and IPL. I don't know how well you know OpenCV, but OpenCV uses the
same image type as IPL, so for many functions
you just have to learn synonyms:  iplCopy(src_ipl_image, dst_ipl_image) should
be written as cvCopy(src_ipl_image, dst_ipl_image) etc.
So you can do 80%-90% of IPL->OpenCV conversion automatically.

Vadim

-----Original Message-----
From: beefheart36 [mailto:testtest36@...]
Sent: Friday, August 01, 2003 5:17 PM
To: OpenCV@yahoogroups.com
Subject: [OpenCV] Old IPL functions in IPP


I have been fool enough to pay for the IPP,
but none of my old IPL functions or variables work e.g. IplImage.

Is there anyway to use these old types or should I port all my code
to use OpenCV functions/types?

Even then how can I be sure that openCV is using the IPP, or even
knows I have it installed?



Change settings: http://www.yahoogroups.com/mygroups, select
    Get Emails (get all posts)
    Daily Digest (one summary email per day)
    Read on the web (read posts on the web only)Or Unsubscribe by mailing
OpenCV-unsubscribe@yahoogroups.com


Your use of Yahoo! Groups is subject to http://docs.yahoo.com/info/terms/

Hello Ayall,

Answers/notices by the items

1) Yes, sure, I have seen it. The document is a bit too low-level, but may be
very useful. Thank you for mentioning it.

2) What you said is true if we use templates internally within the library and
expose only a little part of them outside.
     I meant, if OpenCV had been designed like STL, where data type would be
determined at compile-time, not runtime, that's where
     we could get extremely long compilation times.
     And speaking about library internals, now we are extensively use macros to
handle multiple data types,
     so we hardly can get a better speed than now if we convert them to
templates,
     only more readable code (probably) and more problems with compilers.
     Because most OpenCV users just use OpenCV and do not look at the sources,
     I consider portability more important than readability.  As the compilers
improve and more and more people will like
     to adjust OpenCV to their needs, priorities might change.

3) If someone proposes convenient and efficient object-oriented interface for
any parts of OpenCV, that will be considered.
     For now I think that C++ wrappers for OpenCV data types (IplImage, CvMat,
CvSeq ...) would be certainly useful.

Vadim

-----Original Message-----
From: ayall_gadasy@... [mailto:ayall_gadasy@...]
Sent: Friday, August 01, 2003 5:43 AM
To: OpenCV@yahoogroups.com
Subject: RE: [OpenCV] Re: Writing for Speed


Hi Vadim,

1)
You have mentioned documantations about Pentium up to
3. There is also documantation about Pentium 4. You
might have already seen it, but it may be also good
for others:
http://developer.intel.com/design/pentium4/manuals/248966.htm
It says: "...describes code optimization techniques to
enable you to tune your application for highly
optimized results, when run on Pentium® 4 ..."

2)
You have written that the Templates may increase the
compilation time. That's true, but you do not have to
recompile everything everytime ... You can use the *.h
and *.lib. instead.
BTW, most of the users of the OpenCV use it as a
library (*.h & *.lib) or can do so. Hence they need to
compile the library only once (per version release).
The only case where you need to recompile everything,
is when you change something in a lower level DLL.
Such that all the other parts in the project depends
on it. However, even though you can create a simple
test application that will test your DLL while
debugging your new code. Only then, after the new code
has been verified, you will have to recompile also the
other parts that depends on it.
This way you can save those "hours" of compilation /
recompilation.
I am using Templates, and it is a very powerfull tool.
It's true that it has some restrictions like you have
described. I can add that there are even bugs in the
compiler (MSVC) that you may need to by-pass. (Maybe
they have already been fixed meanwhile.)
Anyway, in my opinion, it is worth to use the
templates. You get a cleaner code, and like Adi said,
you get a better performance!

3)
About the C++ wrappers. It is legitimate to keep using
the old API C function interface, for others. However,
I think that the majority of the OpenCV users, will be
happy to use object-oriented interface.
In these wrapers, you can use inline functions and
transfer the parameters by reference (for example
int&, double& etc.) This way you can get not only a
better interface, but also the same performance, or
almost the same.

Sincerely, Ayall.


--- "Pisarevsky, Vadim" <Vadim.Pisarevsky@...>
wrote:
> Hi Adi,
>
> Answering your questions:
>
> 1. Assembly code in OpenCV
>    There is none. As all the source code is
> available to you, it is easy to check that asm
> (__asm) keyword occurs only once
>    (to raise an exception in cverror.cpp).
>    It's our politics - we provide portable source
> code. If you need top performance, build it with a
> special compiler,
>    or use IPP (directly or indirectly via OpenCV
> interface).
>
> 2. Compiler used to build the binaries.
>     Currently we use MSVC6 and until 6th version of
> Intel Compiler they were comparable.
>     Now ICC7.x seems to outperform MSVC (noticeably
> for some functions, e.g. for gaussian RNG it gives
> ~50% faster code).
>     We will use it to build next OpenCV update.
>
> 3. VectorC. I've seen that once, but there is no
> point for us to use third-party commercial tools
> when we have ICC and IPP.
>     Do you have any links to comparisons VectorC vs.
> ICC or any VectorC benchmarking results?
>
> 4. C++.
>     Indeed, this question has been risen a few
> times.
>     I'm for creating higher-level API for OpenCV,
> but I think, it is important to keep the
>     basic API in C as it is usually simpler to
> create
>     C++, Java, MATLAB, Python etc. wrapper for C
> library than for a set of C++ classes.
>     Anyway, good C++ wrappers for CvMat, IplImage,
> CvSeq etc. are welcome.
>
> 5.  A separate note about templates. C++ compilers
> started to provide more or less complete support for
> templates only recently.
>     Even now there are some bugs, restrictions,
> inconsistencies in template implementation (both
> frontends or backends).
>     For example, MSVC and ICC generate very good
> code, but they are behind GCC in terms of template
> support.
>
>     I would prefer to have a simple code and a good
> compiler able to compile and optimize it correctly,
> rather than a "cool"
>     code that takes few hours of hair pulling
> figuring out where and why the particular compiler
> fails on it.
>
>     There are problems with debugging template
> functions, at least within DevStudio,
>     so here we do not get benefits comparing to
> macros.
>     Templates do not allow to hide implementation
> details that increases
>     compilation time (in some cases up to hours) and
> provide weak encapsulation.
>
>     Templates "infect" the surrounding code, that
> is, a datatype-independent
>     code that uses templates will likely be written
> using templates too.
>     As opposite, with current OpenCV API, which
> provides some
>     sort of RTTI, it is possible to write
> datatype-independent code in C, or any other
> language via a wrapper.
>
>     I think, it will be useful to create a thin
> template-style layer on top of images/matrices (for
> convenient element access)
>     and dynamic structures, such as CvSeq or
> CvGraph, but this is the only part of OpenCV that
> could benefit from templates, IMO.
>
>     Currently we support multiple data types via
> macros and it appears to be quite flexible and
> convenient solution for now - even
>     debugging of the macros is possible after
> half-a-minute preparation effort.
>
>     As for using C++ and templates in IPP - the goal
> of IPP is to provide maximum performance with
> minimum time and code size overhead.
>     Yet, IPP is pretty much flat, i.e. there are
> thousands (literally) of functions of
>     approximately the same (low) level. Having C++
> templates or just a name overloading mechanism on
> top of it makes sense.
>     But notice that in this case we also have C API
> and a thin C++ layer above it, so that both C++ and
> hardcode C developers are satisfied.
>
>     So, as a bottom line, I don't like an idea of
> "switching" API to C++ style, but enhancing OpenCV
> with C++ features on top of existing API
>     (BTW, as you know, we already use optional
> arguments) makes a lot of sense.
>
> Vadim
>
> -----Original Message-----
> From: adishavit [mailto:adish@...]
> Sent: Friday, July 25, 2003 6:53 PM
> To: OpenCV@yahoogroups.com
> Subject: [OpenCV] Re: Writing fo Speed
>
>
> Hi Vadim,
>
>   Thanks for the long and VERY informative reply.
>
> Although I'm familiar (and use) some of the
> techniques, some are new
> to me. I'm sure this will also help others who are
> interested in
> optimizing their code. (A place to start for the
> uninitiated is
> O'Reilly's "High Performance Computing", By Charles
> Severance, Kevin
> Dowd: http://www.oreilly.com/catalog/hpc2/, a little
> old but it has
> all the basics). It is good to know which techniques
> are used by
> OpenCV.
>
> I would love to hear more details when you have the
> time. More
> detailed information regading the implementation of
> each function
> (where relevant) would be great (but a lot of work).
>
> Does OpenCV use any assembly at all?
> Do any of the pre-built DLLs contain assembly
> optimized code, or are
> they just optimized by "better" optimizers/compilers
> (e.g. ICC).
> How do you keep the optimized versions synchronized
> with
> the "regular" C version?
>
> Have you tried VectorC{PC} by www.codeplay.com?
> It is also a vectorizing compiler, which claims to
> be better than
> other compilers in vectorizing C/C++ code.
>
> And finally, the question I have to ask at least
> once a year on this
> list, what about switching to a more C++-ish style?
> Templates can give a lot of compile-time
> optimizations and can make
> the usage of the IPP libraries very simple while
> still keeping things
> like memory handling in-check.
> Constructors, methods and operators would be very
> use (and are) very
> useful for the various data-structures (and little
> more efficient).
> E.g. instead of using cvPoint we could call CvPoint
> ctor.
> Destructor make memory management simpler and less
> painless.
>
> Most of OpenCV users are C++ programmers and most of
> the rest are
> probably using the C subset of C++ with a C++
> compiler.
> Also, most of today's embedded cross-compilers are
> also C++ compilers
> not just C (e.g. gcc, eMVC).
>
> Anyway, sorry about all the questions, I'll be glad
> to see a
> discussion on these topics on the list and not just
> trouble-shooting.
> Excellent post, Vadim, I'll be waiting for more of
> the same :-),
>
> Thanks,
> Adi
>
>
>
>
>
>
>
>
> --- In OpenCV@yahoogroups.com, "Pisarevsky, Vadim"
> <Vadim.Pisarevsky@i...> wrote:
> > Hi Adi,
> >
>
=== message truncated ===


__________________________________
Do you Yahoo!?
Yahoo! SiteBuilder - Free, easy-to-use web site design software
http://sitebuilder.yahoo.com


Change settings: http://www.yahoogroups.com/mygroups, select
    Get Emails (get all posts)
    Daily Digest (one summary email per day)
    Read on the web (read posts on the web only)Or Unsubscribe by mailing
OpenCV-unsubscribe@yahoogroups.com


Your use of Yahoo! Groups is subject to http://docs.yahoo.com/info/terms/

How can I release all the momory allocated for a CvSeq estruture ? As
an example, consider the code below:

void something(IplImage *src)
{
   CvSeq *seq = 0;
   CvMemStorage* storage = cvCreateMemStorage(0);

   cvFindContours( src, storage, &contour, sizeof
(CvContour),CV_RETR_CCOMP, CV_CHAIN_APPROX_SIMPLE );

   cvReleaseMemStorage(&storage);

  // How can I release completely the memory used by contour ?

}

Regards

Andres

You can use the CvtColor function available in OpenCv. See the HTML
doc. for more detail. Alternatively, you can use more robust technics
like 3D clustering of the RGB space or neural networks were the input
can be the owner RGB values, or a combination of RGB + H (HSV).
However, if your problem is simple, try first the simple approach of
HSV space.

good luck

andres

--- In OpenCV@yahoogroups.com, farooq azam <farooqsinn@y...> wrote:
> Hi all!
>
> I don't have much concept of computer  vision. I have
> a simple prob I hope someone can help me out.
>
> I have an RGB image of an object in two different
> lightning conditions. I would like to recognize the
> object on tha basis of its color (object has only one
> color). My basic recognition program doesn't work on
> these.
>
> I read somewhere that HSI (Hue, saturation &
> intensity)image can cater intensity changes better. Is
> HSI implemented in openCV or would HSI be better? how
> do we create an HSI image.
>
> Regards
> Farooq
>
>
>
> __________________________________
> Do you Yahoo!?
> Yahoo! SiteBuilder - Free, easy-to-use web site design software
> http://sitebuilder.yahoo.com

Sorry,
I know this does not answer your question but this link may be
interesting.

http://w3.impa.br/~pcezar/3dp/original/CVL_html/appPage/calib_doc/para
meters.html

it talks a little about the matlab version of this calibration
software, I believe it is identical.
More specifically, (near the end) it talks about the calibration
plate coordinate system/reference frame.

it is interesting to note that the y-axis is positive downwards in
the calibration frame does this require the image plane to also have
y +ve down?

mmm not sure, I guess it does?

JA

--- In OpenCV@yahoogroups.com, "melanie_vaillant"
<melanie_vaillant@y...> wrote:
> Hello all!
>
> Here is my problem:
>
> I have calibrated a camera, and I am capturing some frames from it.
>
> Now, if I know the coordinates of a point m on the floor, in the
> image plan, how can I get the coordinates of the same point, M, on
> the real floor?
>
> As I have my camera parameters, I have the intrinsic matrix A, the
> rotation matrix R and the translation vector t, so I know the
> transformation:
>                        m = A(RM+t) (<=> m-At = ARM)
> Lets P = AR and b = m-AT.
>
> In my opinion, finding M knowing m is equivalent to solve the
> equation: PM=b.
> I have done this, using a LU decomposition.
>
> But, I mean that all points of the line (M m) have the same image m
> in the image plane.
> So, as my computer find one and only one solution, how can I find
the
> point on the real foor that have m as image?
>
> Is anybody could help me?
>
> Many thanks for any answer,
>
> Mélanie

Hi Andres,

>  // How can I release completely the memory used by
> contour ?

You have just did it ... when you used:
>   cvReleaseMemStorage(&storage);

Sincerely, Ayall.


--- aperez_br <aperez_br@...> wrote:
> How can I release all the momory allocated for a
> CvSeq estruture ? As
> an example, consider the code below:
>
> void something(IplImage *src)
> {
>   CvSeq *seq = 0;
>   CvMemStorage* storage = cvCreateMemStorage(0);
>
>   cvFindContours( src, storage, &contour, sizeof
> (CvContour),CV_RETR_CCOMP, CV_CHAIN_APPROX_SIMPLE );
>
>   cvReleaseMemStorage(&storage);
>
>  // How can I release completely the memory used by
> contour ?
>
> }
>
> Regards
>
> Andres
>
>


__________________________________
Do you Yahoo!?
Yahoo! SiteBuilder - Free, easy-to-use web site design software
http://sitebuilder.yahoo.com

cvReleaseMemStorage(&storage);
Is this command sufficient to release all the memory ? Including the
momery used by the CvSeq *contour ? For example, the code below,
finds the major binary object in an image. Some part of this code,
whose I don't find it yet, allocate memory without release it. If
someone can help me, I will be very glad.

andres

//this method finds the major object in a binay image
CvRect __fastcall imagem::IFindMObject(IplImage *image)
{
    CvMemStorage* storage = cvCreateMemStorage(0);
    IplImage *dst = cvCreateImage( cvSize(image->width,image->height),
8, 1 );

    CvSeq  *contour = 0;
    CvSeq *C_maior=0;

    CvRect bound,Rmaior;
    double maior=0;
    int size;
    int color;

    dst=cvCloneImage(image);
    //Finds each image object through connected components
    cvFindContours( dst, storage, &contour, sizeof(CvContour),
CV_RETR_CCOMP, CV_CHAIN_APPROX_SIMPLE );

    //Clean the destination image with black
    cvZero( dst );

    for( ; contour != 0; contour = contour->h_next)
      {
          bound=cvBoundingRect(contour,0); //Finds the bounding
          cvSetImageROI(image,bound);    //Calculates the image area.
          size=cvCountNonZero(image);

          if(size>maior)
             {
                maior=size; //stores the area of the major object and
                             //and point to it
                C_maior=contour;
                Rmaior=bound;
             }
      }

    cvResetImageROI(image);  //Resets the input function image
    cvZero(image);      //clean the image with black
    color = CV_RGB(255,255,255);
    //Draws the major object in the image
    cvDrawContours( image, C_maior, color, color, -1, CV_FILLED, 8 );

    //release image memory
    cvReleaseImage(&dst);

    //release memory alocated for CvSeq structure
    cvReleaseMemStorage(&storage);

    return(Rmaior);
}
//-------------------------------------------------------------------




Thanks

andres

--- In OpenCV@yahoogroups.com, <ayall_gadasy@y...> wrote:
> Hi Andres,
>
> >  // How can I release completely the memory used by
> > contour ?
>
> You have just did it ... when you used:
> >   cvReleaseMemStorage(&storage);
>
> Sincerely, Ayall.
>
>
> --- aperez_br <aperez_br@y...> wrote:
> > How can I release all the momory allocated for a
> > CvSeq estruture ? As
> > an example, consider the code below:
> >
> > void something(IplImage *src)
> > {
> >   CvSeq *seq = 0;
> >   CvMemStorage* storage = cvCreateMemStorage(0);
> >
> >   cvFindContours( src, storage, &contour, sizeof
> > (CvContour),CV_RETR_CCOMP, CV_CHAIN_APPROX_SIMPLE );
> >
> >   cvReleaseMemStorage(&storage);
> >
> >  // How can I release completely the memory used by
> > contour ?
> >
> > }
> >
> > Regards
> >
> > Andres
> >
> >
>
>
> __________________________________
> Do you Yahoo!?
> Yahoo! SiteBuilder - Free, easy-to-use web site design software
> http://sitebuilder.yahoo.com

Hi, Andres!
I don't think it's needed to use cvCreateImage and then cvCloneImage
for the dst pointer. Just us only cvCloneImage.
"From the OpenCV documentation:
CloneImage
Makes a full copy of image

IplImage* cvCloneImage( const IplImage* image );

image
Original image.

The function cvCloneImage makes a full copy of the image including
header, ROI and data "

Hope it would help,
Chris


--- In OpenCV@yahoogroups.com, "aperez_br" <aperez_br@y...> wrote:
> cvReleaseMemStorage(&storage);
> Is this command sufficient to release all the memory ? Including
the
> momery used by the CvSeq *contour ? For example, the code below,
> finds the major binary object in an image. Some part of this code,
> whose I don't find it yet, allocate memory without release it. If
> someone can help me, I will be very glad.
>
> andres
>
> //this method finds the major object in a binay image
> CvRect __fastcall imagem::IFindMObject(IplImage *image)
> {
>    CvMemStorage* storage = cvCreateMemStorage(0);
>    IplImage *dst = cvCreateImage( cvSize(image->width,image-
>height),
> 8, 1 );
>
>    CvSeq  *contour = 0;
>    CvSeq *C_maior=0;
>
>    CvRect bound,Rmaior;
>    double maior=0;
>    int size;
>    int color;
>
>    dst=cvCloneImage(image);
>    //Finds each image object through connected components
>    cvFindContours( dst, storage, &contour, sizeof(CvContour),
> CV_RETR_CCOMP, CV_CHAIN_APPROX_SIMPLE );
>
>    //Clean the destination image with black
>    cvZero( dst );
>
>    for( ; contour != 0; contour = contour->h_next)
>      {
>          bound=cvBoundingRect(contour,0); //Finds the bounding
>          cvSetImageROI(image,bound);    //Calculates the image
area.
>          size=cvCountNonZero(image);
>
>          if(size>maior)
>             {
>                maior=size; //stores the area of the major object
and
>                             //and point to it
>                C_maior=contour;
>                Rmaior=bound;
>             }
>      }
>
>    cvResetImageROI(image);  //Resets the input function image
>    cvZero(image);      //clean the image with black
>    color = CV_RGB(255,255,255);
>    //Draws the major object in the image
>    cvDrawContours( image, C_maior, color, color, -1, CV_FILLED,
8 );
>
>    //release image memory
>    cvReleaseImage(&dst);
>
>    //release memory alocated for CvSeq structure
>    cvReleaseMemStorage(&storage);
>
>    return(Rmaior);
> }
> //-----------------------------------------------------------------
--
>
>
>
>
> Thanks
>
> andres
>
> --- In OpenCV@yahoogroups.com, <ayall_gadasy@y...> wrote:
> > Hi Andres,
> >
> > >  // How can I release completely the memory used by
> > > contour ?
> >
> > You have just did it ... when you used:
> > >   cvReleaseMemStorage(&storage);
> >
> > Sincerely, Ayall.
> >
> >
> > --- aperez_br <aperez_br@y...> wrote:
> > > How can I release all the momory allocated for a
> > > CvSeq estruture ? As
> > > an example, consider the code below:
> > >
> > > void something(IplImage *src)
> > > {
> > >   CvSeq *seq = 0;
> > >   CvMemStorage* storage = cvCreateMemStorage(0);
> > >
> > >   cvFindContours( src, storage, &contour, sizeof
> > > (CvContour),CV_RETR_CCOMP, CV_CHAIN_APPROX_SIMPLE );
> > >
> > >   cvReleaseMemStorage(&storage);
> > >
> > >  // How can I release completely the memory used by
> > > contour ?
> > >
> > > }
> > >
> > > Regards
> > >
> > > Andres
> > >
> > >
> >
> >
> > __________________________________
> > Do you Yahoo!?
> > Yahoo! SiteBuilder - Free, easy-to-use web site design software
> > http://sitebuilder.yahoo.com

hi,
To compute the eigenface, I use the opencv's eigenobject function,
icvCalcEigenObjects_8u32fR (int nObjects,
                             void*  input,
                             int    objStep,
                             void*  output,
                             int    eigStep,
                             CvSize size,
                             int    ioFlags,
                             int    ioBufSize,
                             void*  userData,
                             CvTermCriteria* calcLimit,
                             float* avg,
                             int    avgStep,
                             float  *eigVals)

but I don't know what the function's parameters mean (so many
parameters).

So I seek another way, first I want to compute the covariance matrix,
then use icvJacobiEigens_32f (float *A,
                               float *V,
                               float *E,
                               int n,
                               float eps)
But I don't how to use the icvCalcCovarMatrixEx_8u32fR(int nObjects,
                                                       void *input,
                                                       int objStep1,
                                                       int ioFlags,
                                                       int ioBufSize,
                                                       uchar * buffer,
                                                       void *userData,
                                                       float *avg,
                                                       int avgStep,
                                                       CvSize size,
float *covarMatrix) to compute the covariance matrix.

if anyone has a simple way to compute the eigenface, could you please
send email to me? thanks.

I can't get my project to compile with the CV Libraries. I keep getting the "unresolved external symbol" error for every function I try.

When I try to run the minarea.c sample y obtain:

Linking...

minarea.obj : error LNK2001: unresolved external symbol _cvDestroyWindow

minarea.obj : error LNK2001: unresolved external symbol _cvWaitKey

minarea.obj : error LNK2001: unresolved external symbol _cvShowImage

minarea.obj : error LNK2001: unresolved external symbol _cvLine

minarea.obj : error LNK2001: unresolved external symbol _cvCircle

minarea.obj : error LNK2001: unresolved external symbol _cvSetZero

minarea.obj : error LNK2001: unresolved external symbol _cvBoxPoints

minarea.obj : error LNK2001: unresolved external symbol _cvMinEnclosingCircle

minarea.obj : error LNK2001: unresolved external symbol _cvMinAreaRect2

minarea.obj : error LNK2001: unresolved external symbol _cvNamedWindow

minarea.obj : error LNK2001: unresolved external symbol _cvCreateImage

Debug/minarea.exe : fatal error LNK1120: 11 unresolved externals

Error executing link.exe.

minarea.exe - 12 error(s), 0 warning(s)

I have included all the following INCLUDE directories in the include files

C:\….\OPENCV\OTHERLIBS\_FLTK\INCLUDE

C:\….\OPENCV\OTHERLIBS\_GRAPHICS\INCLUDE

C:\….\OPENCV\OTHERLIBS\HIGHGUI

C:\….\OPENCV\CV\INCLUDE

And the LIB directories in the library files,

C:\….\OPENCV\LIB

C:\….\OPENCV\OTHERLIBS\_GRAPHICS\LIB
C:\….\OPENCV\OTHERLIBS\_FLTK\LIB

And finally included in the Executable files,

C:\….\OPENCV\BIN

hello,

I have min/max focus and the min/max angles of view of my camera and
some other intrinsic parameters (lens,...).
How can calculate the angle of view for a focus value between its
min/max focus? What is the optical system model for this?
I saw in OpenCV documentation a function for camera calibration but it
seems not to contain the formulas i need...

thanksfor any pointers/hints!
--florian

I'm trying to compile OpenCV with Visual Studio .NET 7, Direct X SDK
9, and openCV 3.1b on a machine running XP.

I've compiled the directshow baseclasses and added the Debug and
Release directories to the library search path. I added the
baseclasses folder to the include path.

Here are my include directories:
C:\DXSDK\Samples\C++\DirectShow\BaseClasses
C:\DXSDK\Include
C:\tk8.4.4\xlib
C:\tk8.4.4\generic
C:\tcl8.4.4\generic
C:\Program Files\OpenCV\otherlibs\highgui
C:\Program Files\OpenCV\cvaux\include
C:\Program Files\OpenCV\cv\include
C:\Program Files\Matrox Imaging\mil\include
.
.
.

Here are my lib directories:
C:\DXSDK\Samples\C++\DirectShow\BaseClasses\Release
C:\DXSDK\Samples\C++\DirectShow\BaseClasses\Debug
C:\DXSDK\lib
C:\tk8.4.4\win\Release
C:\tcl8.4.4\win\Release
C:\Program Files\OpenCV\otherlibs
C:\Program Files\OpenCV\lib
C:\Program Files\Matrox Imaging\mil\library\winnt\msc\dll
.
.
.

I've also built tcl/tk and I have MIL-Lite 7.5 installed.

I keep getting DirectX linker and compiler errors:
C:\DXSDK\Include\ddraw.h(231): error C2143: syntax error : missing ';'
before 'identifier'

C:\DXSDK\Include\ddraw.h(231): fatal error C1075: end of file found
before the left brace '{' at 'C:\DXSDK\Include\ddraw.h(48)' was matched

C:\DXSDK\Include\ddraw.h(231): warning C4091: 'typedef ' : ignored on
left of '_DDARGB' when no variable is declared

c:\Program Files\Microsoft Visual Studio .NET
2003\Vc7\include\varargs.h(90): warning C4005: 'va_start' : macro
redefinition

c:\Program Files\Microsoft Visual Studio .NET
2003\Vc7\include\varargs.h(92): warning C4005: 'va_end' : macro
redefinition

c:\Program Files\Microsoft Visual Studio .NET
2003\Vc7\PlatformSDK\Include\MMSystem.h(146): error C2143: syntax
error : missing ';' before 'identifier'

c:\Program Files\Microsoft Visual Studio .NET
2003\Vc7\PlatformSDK\Include\MMSystem.h(146): fatal error C1075: end
of file found before the left brace '{' at 'c:\Program Files\Microsoft
Visual Studio .NET 2003\Vc7\PlatformSDK\Include\MMSystem.h(42)' was
matched

c:\Program Files\Microsoft Visual Studio .NET
2003\Vc7\PlatformSDK\Include\MMSystem.h(146): warning C4091: 'typedef
' : ignored on left of 'mmtime_tag' when no variable is declared

etc.. etc...
and then at the end a bunch of things like this:
SyncFilter fatal error LNK1120: 37 unresolved externals
SyncFilter error LNK2001: unresolved external symbol _MEDIASUBTYPE_RGB32
SyncFilter error LNK2001: unresolved external symbol _MEDIASUBTYPE_RGB4
SyncFilter error LNK2001: unresolved external symbol _MEDIASUBTYPE_RGB555
SyncFilter error LNK2001: unresolved external symbol _MEDIASUBTYPE_RGB565
SyncFilter error LNK2001: unresolved external symbol _MEDIASUBTYPE_RGB8
SyncFilter error LNK2001: unresolved external symbol _MEDIATYPE_Audio
SyncFilter error LNK2001: unresolved external symbol _MEDIATYPE_Audio
SyncFilter error LNK2001: unresolved external symbol _MEDIATYPE_Video


In total there are 1397 build errors. I'm trying to build the MIL
Debug version. I get the same errors with any of the other
distributions, though.

I got it to compile on another XP machine. However, I had to install
VC.Net on this machine myself. Could it be misconfigured or
something?? HELP, please!

Thanks,
Reuben Grinberg
reuben.grinberg@...

Hi,all.
   I want someone can provide me with some code about how to read the video frame
data from a yuv
file(cif or qcif);

   like a c function in this way:
   void Get_YUV_Frame(int *Y,int *U,int *V,char *yuv_filename,long frame_number)
   {
       //To get the no.frame_number frame video data into y,u,v from
yuv_filename.
       //assume it's a qcif file
       //so Y is a matrix of 176x144;
       //   u,v, is a matrix of 88x72;
   }

    thanks a lot!
    your reply will be very helpful to me!



=====
Regards
Yan Jiangguo

contact:
koogio@...
http://wavelet.nease.net/yan

__________________________________
Do you Yahoo!?
Yahoo! SiteBuilder - Free, easy-to-use web site design software
http://sitebuilder.yahoo.com

hi

When using 2 images of the same object but with different rotations
the function Hu Moments is not finding the same moments.  The
moments are so different.  Why is this happening??

Anyone has an example of the correct use of the function??

Im using binary images with no ROI.  I want to apply Hu moments to
the whole image.

IplTemp is a 1 channel image.

My code is:


         cvMoments(IplTemp,&momentos,0);

         cvGetHuMoments(&momentos,&humomentos);

         vect[0]=humomentos.hu1;
         vect[1]=humomentos.hu2;
         vect[2]=humomentos.hu3;
         vect[3]=humomentos.hu4;
         vect[4]=humomentos.hu5;
         vect[5]=humomentos.hu6;
         vect[6]=humomentos.hu7;



Thanks

Millan Sanchez

hi all
i want to load a 10-bit per pixel bitmap into 8-bit per channel ipl image
i red about iplConvertFromDIB() but it need BITMAPINFOHEADER*
i tried using CBitmap with LoadImage() but i couldn't find any BITMAPINFOHEADER

can u help me?

thanx

hi all
i want to load a 10-bit per pixel bitmap into 8-bit per channel ipl image
i red about iplConvertFromDIB() but it need BITMAPINFOHEADER*
i tried using CBitmap with LoadImage() but i couldn't find any BITMAPINFOHEADER

can u help me?

thanx

16bpp instead of 10bpp

hi all
i want to load a 16-bit per pixel bitmap into 8-bit per channel RGB ipl image
i red about iplConvertFromDIB() but it need BITMAPINFOHEADER*
i tried using CBitmap with LoadImage() but i couldn't find any BITMAPINFOHEADER

can u help me?

thanx

Hi Mohamed Zayed,
Sorry for replying late. I didn't see ur message untill recently.
Thanks for ur tips first.

I still having problem about the result i got from the function
cv3dTrackerCalibrateCameras. All numbers i got still not making any
sence to my set up.

This is typical result i got from my program, camera_info[0].mat

CvBool
                  ☺
Matrix: -- maps camera coordinates to world coordinates

[   -0.00199   0.000769 -7.23e-005          0 ]
[   0.000719    0.00192   0.000616          0 ]
[     -0.134     -0.256      0.957          0 ]
[       1.36      -98.9        446          1 ]

principal_point          168.80, 148.35

As far as i understand, the last line for example " 1.36  -98.9
446  " is supposed to tell the position of this camera in 3d
coordinates referring to the (0,0,0) point which is the left and top
corner of the chessbroad. And everything should be in the same scale
as the square size that u inputed for the funtion. I am using mm.

I just don't know where i did anything wrong.
There are only these a few parameters for this function. Do u know
where could be wrong ? Most of them are typical values.


cv3dTrackerCalibrateCameras(
		 2,       //number of cameras
		 intrinsics, //const cameraIntrinsics
		 cvSize(6,6 ), //checkerboard_size
		 (float) 25.4 ,    //square size ,mm
		 stereo_samples , // a view from each camera
		 camera_info );

What u say ? Is there anything else i should pay attention ? Or is
there special way to set up the stereo ? Would u show me some result
u got ? And what do u think about the function in terms of
accuracy ?
Thanks.
Cheers.

Eric



--- In OpenCV@yahoogroups.com, "Mohamed Zayed" <zayed@e...> wrote:
> Hi Eric,
> Well, I tested these functions a long time ago but I have to work
on it to
> exploit them for my applacation.
> Anyway, I will try to respond at some of your questions :
> 1- The reference of the world is the reference of the first
chessboard " the
> top left corner of the first chessboard image"
> thus, any measure of 3D locations will be done in this reference.
( To be
> sure, try to measure a distance between to chessboard corners in
the first
> image).
> 2- The structures used in these functions uses calibration
parameters in an
> elegant way ( so its juste a good representation of camera
parameters to
> simplify the usage).
>
> Good luck
> Mohamed ZAYED
>
> ----- Original Message -----
> From: "eric_qlink" <eric_qlink@y...>
> To: <OpenCV@yahoogroups.com>
> Sent: Sunday, July 13, 2003 11:14 PM
> Subject: [OpenCV] Question about 3D Tracking Functions
>
>
> > Hi there,
> > I have been doing some work about stereo calibration.
> > I tried openCV 3D Tracking Functions for stereo calibration and
3d
> > tracking, like
> > 3dTrackerCalibrateCameras
> > 3dTrackerLocateObjects
> >
> > But i still don't understand the result data i got from them.
> > For example :
> > how the Cv3dTrackerCameraInfo maps the position of two cameras
> > related to the chessborad in the scene. What the 4x4 matrix
means ?
> > And what does the Cv3dTrackerTrackedObject tell us ? The 3d
position
> > of the target. But where is the reference coordinate ? The
center of
> > the chessborad ?
> >
> > I couldn't find enough discription about the funtions.
> > So wondering if anybody got some exprience about it. Please help
me
> > out.  Thanks in advance.
> >
> > Cheers.
> >
> > Eric
> >
> >
> >
> >
> > Change settings: http://www.yahoogroups.com/mygroups, select
> >    Get Emails (get all posts)
> >    Daily Digest (one summary email per day)
> >    Read on the web (read posts on the web only)Or Unsubscribe by
mailing
> OpenCV-unsubscribe@yahoogroups.com
> >
> >
> > Your use of Yahoo! Groups is subject to
http://docs.yahoo.com/info/terms/
> >
> >

if the image is on the hard drive you could try
IplImage* cvLoadImage( const char* filename, int iscolor CV_DEFAULT
(1));

The BITMAPINFOHEADER is a windows structure which describes the
bitmap,  but I would stay away from that stuff if you can its a real
pain.

I havent tried cvLoadImage myself but it looks easy to use.

Also iplConvertFromDIB() seems to have been removed from the new IPP3

JA



--- In OpenCV@yahoogroups.com, "Mohamed Galal Eldin" <eng_mg@y...>
wrote:
> 16bpp instead of 10bpp
>
> hi all
> i want to load a 16-bit per pixel bitmap into 8-bit per channel RGB
ipl image
> i red about iplConvertFromDIB() but it need BITMAPINFOHEADER*
> i tried using CBitmap with LoadImage() but i couldn't find any
BITMAPINFOHEADER
>
> can u help me?
>
> thanx

Hello everybody...

Hope tthat anybody can help me...

The Problem: I have to write a function,that reads out the pixel-
information of an Ipl-Image and than putds them into a two dimensional
array. how can I do this?

Cheers
Alex

Probably the problem is that you have to specify a background file,
that createsamples uses to create num=1000 object images with
different background.
It could be useful to specify also maxidev, maxanglex etc to make
your samples as general as possible.
Bye!

Valentina

P.S.: how can I open positive.vec file?



--- In OpenCV@yahoogroups.com, "Matthias Wimmer" <yahoo@w...> wrote:
> Hi,
>
> I have problems with createsamples.exe:
> This is my command line:
>
> createsamples -img positives\auge002.bmp -vec data\positives.vec -
num
> 1000 -w 50 -h 50 -show
>
>
> A window appears on the screen that is intended to show all the
> generated images.
>
> That is where the problem is located: All the images are black. And
> the file data\positives.vec is quite homogenous.
>
>
> Please give me some advice
>
> Matt

in order to access pixels have a look at the pixel access macros:


http://groups.yahoo.com/group/OpenCV/message/13433


Matt





> -----Original Message-----
> From: alex_krahl [mailto:Alex_Krahl@...]
> Sent: Monday, August 04, 2003 10:29 AM
> To: OpenCV@yahoogroups.com
> Subject: [OpenCV] Problem with converting.... Please Help!!!
>
>
> Hello everybody...
>
> Hope tthat anybody can help me...
>
> The Problem: I have to write a function,that reads out
> the pixel-
> information of an Ipl-Image and than putds them into a
> two dimensional
> array. how can I do this?
>
> Cheers
> Alex
>
>
> ------------------------ Yahoo! Groups Sponsor
>
> Change settings: http://www.yahoogroups.com/mygroups, select
>    Get Emails (get all posts)
>    Daily Digest (one summary email per day)
>    Read on the web (read posts on the web only)Or
> Unsubscribe by mailing OpenCV-unsubscribe@yahoogroups.com
>
>
> Your use of Yahoo! Groups is subject to
http://docs.yahoo.com/info/terms/