nguyenducnhaty / aforge Goto Github PK

The Stevenson-Arce dithering method is supposed to work with
hexagonal pixel lattices. This is why half the elements in the
matrix are null, which gives very poor results in AForge.

My opinion is that it should be completely removed from AForge,
because it gives people the wrong idea that the Stevenson-Arce
algorithm is mediocre (on the intended support, it isn't).

Hi Andrew

Would there be much work involved in adding the ability to capture frames 
from a mms stream in teh same way that you can from local devices using 
direct show  etc 

Thanks

Karl

It is required to implement Homomorphic filtering.

Useful links:
http://en.wikipedia.org/wiki/Homomorphic_filtering
http://www.cs.sfu.ca/~stella/papers/blairthesis/main/node35.html

In polish postfix expressions it is required to partly (or fully) evaluate 
expression variables - replace variables with their values (for example 
replace $2 variable with its value equal to 3.14)

What steps will reproduce the problem?
1. Open iplab
2. Click on the show histogram icon from the toolbar.
3. Now open any image
4. Close the histogram viewer.
5. Close the application.

What is the expected output? What do you see instead?
The program should exit. We get an error message say "Cannot access a
disposed object".

What version of the product are you using? On what operating system?
I am using v2.4.0 on windows XP professional. 

Please provide any additional information below.
The screen shot and the details are available in the attached file.

Not really a problem, just a missing feature...

Assembly: AForge.Video.DirectShow
Class: VideoCaptureDevice
Version 1.6.0

It would be interesting to be able to set some basic parameters related to 
the video capture device, like the frame rate, the video size 
(resolution), brightness, contrast,, signal type (RGB24, i420, etc) and so 
on. WDM devices usually provide these functions.

At this point ComplexImage class keeps complex image's data privately. It 
is required to provide access to them, so it would be possible to develop 
additional filters.

It would be nice if the framework could provide a built-in way to validade 
neural networks based on a validation set. The sample code provided below 
may be useful to accomplish this task:

//AForge.Neuro, in class BackpropagationLearning.cs:

<code> 
        public double MeasureError(double[] input, double[] output) 
        { 
            // compute the network's output 
            network.Compute(input); 

            // calculate network error 
            double error = CalculateError(output); 

            return error; 
        } 

        public double MeasureError(double[][] input, double[][] output) 
        { 
            double error = 0.0; 

            for (int i = 0; i < input.Length; ++i) 
            { 
                error += MeasureError(input[i], output[i]); 
            } 

            return error; 
        } 
</code>

if you remember we discussed it before and you said that removing locks
will increase the performance i think that this is the solution for that

What steps will reproduce the problem?
1. make methods which directly inputs Bitmapdata
2. make class which handles locking and unlocking of images

What is the expected output? What do you see instead?
faster application and image processing

What version of the product are you using? On what operating system?
[13.06.2006] - Version 2.3.0 winXP

Implementation of a Template Matching algorithms would be a good plus for 
the library.
To do that some function for global minimum/maximum detection over a 2d 
image would also be required.

At this point AForge.Imaging supports 24 bpp RGB and 8 bpp indexed 
(grayscale) images. It would be nice to extend the library and support 32 
bpp (RGBA) images as well.

This library is infinitely useful and I've had some limited success with
using the pre-compiled binaries in VB.NET, however I get errors when I
attempt to use the BlobCount functions. Some simple VB.NET code samples
would be great for those who find C# a little too daunting.

It is required to have a possibility to compile polish postfix expression 
to IL code, so it could be called faster.

does aforge has following functionality

1) automatically crop the image
2) auto deskew the image 
4) hide micr line on check image.

if yes then how to implement 

thanks

What steps will reproduce the problem?
1. Compile and run the Time Series example with Visual Studio 2005.
2. You will get the run-time error: "Cross-thread operation not valid: 
Control 'currentIterationBox' accessed from a thread other than the thread 
it was created on."

Under VS 2005 you cannot run code that is not thread safe.
See: http://msdn2.microsoft.com/en-us/library/ms171728.aspx for more 
details.

What version of the product are you using? On what operating system?
Windows XP, Visual Studio 2005.

In the case of big images (around 4000x3000) images statistics calculates 
incorrect mean value, which is negative.

At this point current motion detection classes are based on the two frames 
difference or on background modeling. In the case if new method for motion 
frame calculation is created, it will be required to create new motion 
detector. It is required to split motion detection classes to a set of 
classes, which calculate motion frames (binary frame with motion regions), 
and set of classes, which process the motion frame further.

At this point BlobCounterBase class allows to retrieve all objects’ 
rectangles (GetObjectRectangles()) or all objects (GetObjects()). But 
there is not way to extract particular object (there is no way also to 
reference it).

Use case. For example, using GetObjectRectangles() we retrieve all 
rectangles and then, after their processing, we need to extract only the 
biggest blob. GetObjects() will give entire objects’ collection, but not 
just one blob, what is not efficient from performance stand point.

Add mirror filter to AForge.Imaging library. The filter should be able to 
mirror using X and/or Y axis.

Could anyone give me some pointers on how to capture images at 640x480 
instead of teh default 320x240 reoslution?

Many thanks,

It is required to develop facility, which allows applying image processing 
filter to part of source image only.

In DoubleRange.cs and IntegerRange.cs, I think the IsInside function is in
error; Don't really know how it affect the code and examples, I'm just
starting to look at it !

Original:
public bool IsInside( double x )
{
  return ( ( x >= min ) && ( x <= min ) ); // This looks like return x == min
}

Should be (?) :
  return ( ( x >= min ) && ( x <= max ) );

It is required to simplify polish postfix expressions – evaluate 
functions, which have constants as their arguments, and replace these 
functions with their result.

Some video files, which can be played fine using WMP, can be played using 
classes of AForge.Video.DirectShow - failed connection filters error is 
provided.

Hi, I'm working in a project called GeoMetriX, it works with a camera that 
i projected to an area of 40cm X 40cm, in the area i have circles, squares 
and triangles, i already binarize the image and apply a sobel edge 
detector, so i already have only the edges of the figures.
I only have 3 circles 3 squares and 3 triangles at the moment, but when i 
used Hough circles detector with 
AForge.Imaging.HoughCircleTransformation HCT = new 
AForge.Imaging.HoughCircleTransformation(4);
OR
AForge.Imaging.HoughCircleTransformation HCT = new 
AForge.Imaging.HoughCircleTransformation(5);
It detects like 234 circles, mmm i dont know if i should change the 
intensityt min and max, or what i should do. 

Thanks, and sorry for my english, im from Mexico.

Aldo Sandoval Monsivais
[email protected]

It would be nice to have support of 48 and 64 bpp color images.

At this point most AForge.Imaging filters support filtering on Bitmap or 
BitmapData. It would be nice to extend the library to support raw images 
in unmanaged memory. This will give possibility to do more optimized image 
processing in some cases and potentially may give compatibility with some 
other libraries (Paint.NET ?).

It is required to implement population shuffling, when new generation is 
generated. Now the feature is implemented in Elite Selection algorithm. It 
is better to move the functionality to Population class and allow to 
enable/disable the function.

The following is a code i wrote to test the Complex object in Aforge.Math. 
Complex[] dx = new Complex[512];
for (int i = 0; i < dx.Length-1; i++)
{
dx[i] = (Complex)(Math.Sin(i));//ERROR????
}
When I run this, the compiler(Visual studio.NET 2005) complains"Cannot 
convert type double to Aforge.Math.Complex".
Has anyone ever encountered this error.

Can anyone upload some more examples(tutorial) on the usage of Math 
functions in this library.
Cheers 


kel

What steps will reproduce the problem?

Bitmap _bmpToFilter = new Bitmap.LoadFromFile(fileName);
Bitmap _bmpT = new Threshold().Apply(_bmpToFilter);

What is the expected output? What do you see instead?

The Threshold().Apply() method is overloaded and accepts also Bitmaps or
BitmapData but only works if you pass a BitmapData like this...

IFilter grayscaleFilter = new GrayscaleRMY();
Bitmap _bmpToFilter = new Bitmap.LoadFromFile(fileName);
Bitmap _bmpT = new Threshold().Apply(grayscaleFilter.Apply(_bmpToFilter));


What version of the product are you using? On what operating system?

Version 1.5.1 on Windows XP Professional SP2
NET Framework 2.0

Please provide any additional information below.

I need to save trained network to file. The best way how to do it is 
enable serialization Network class.

Hi Andrew,

Congratulations on your excellent work on Aforge, specifically Image 
Processing!

It would be great if you could add features specifically on snakes or 
active contour.  

Jason Xie has a good paper on this website... 
http://www.cs.bris.ac.uk/Research/Vision/AC/

It would be nice to add a filter, which extracts the biggest found blob on 
the image.

Objects found by Blob Counter should be ordered, but not in the "random" 
order, which is determined by algorithm itself.

Would be great to have a way to provide minimum and maximum sizes for blob
detection to the BlobCounter class in order to reduce client code in the
applications.

Perhaps two properties of class Size can be provided (i.e. BlobMaxSize and
BlobMinSize) and if they are not filled by the user, the size check not be
internally performed.

Usage VB sample code:

thisBlobCounter.BlobMinSize(new Size(20,30))
thisBlobCounter.BlobMaxSize(new Size(80,120))

Sandro

It would be nice to have posibility to convert image from RGB to Lab color 
mode and vice-versa.

It is required to implement Harris corner detector. Some information can 
be found here: 
http://en.wikipedia.org/wiki/Corner_detection#The_Harris_.26_Stephens_.2F_P
lessey_corner_detection_algorithm

It would be nice to have motion detector, which does not do any background 
modeling on its own, but uses predefined scene image. First image from 
video sequence may be used as scene image, if it was not set before. User 
should be able specify scene image at any time before or after motion 
detection was started.

Elite selection should randomly sort new population, so it will not 
contain all best chromosomes on the top.

When getting video from a local video capture device using the
AForge.Video.DirectShow classes if the display changes (screen saver kicks
in or CTRL+ALT+DEL is pressed) then the video stops updating.

Can test this using the motion detection example.

Think the EC_DISPLAY_CHANGED messages may need to be processed but not sure.

Hi,

is there a way to use aforge under vb.net?

Thanks

What steps will reproduce the problem?
1. OilPainting.Apply(Bitmap)
2.
3.

What is the expected output? What do you see instead?
Returned Bitmap is usually only partially filtered (20% - 40%).
The remaining portion is solid black.

What version of the product are you using? On what operating system?
1.4.2
XP
Vista

Please provide any additional information below.

 void OilPaintingNode_DoProcessing(object sender)
 {
     //process image if xinput contains data
     if (xinput.Data != null)
     {
         OilPainting oip = new OilPainting();

         //set brush size if bsize contains data
         if (bsize.Data != null)
         {
             oip.BrushSize = (int)bsize.Data;
         }

         outpin.Data = oip.Apply((Bitmap)xinput.Data);

         if (UINode != null)
         {
             UINode.BackgroundImage = (Bitmap)outpin.Data;
         }
     }
 }

It is required to recalculate fitness value only of new members of the 
population. If input data don't change during solution search, we do not 
need to recalculate fitness of the entire population, what leads to 
decreasing performance.

== Situation ==
Many samples such as Neuro\Back Propagation\Approximation read a .csv file
and parse decimal numbers, which are formatted with a colon '.' as the
decimal separator.

== Problem ==
In the regional settings of many countries (e.g. most European countries,
Russia, South America...), a comma ',' is used as a decimal separator, and
not a colon '.'
In those countries, the samples parsing decimal numbers fail to load the
.csv file.

== Solution ==
Use for instance CultureInfo.InvariantCulture when parsing those numbers.

=== Code ===
using System.Globalization;
...
double.Parse( strs[0], CultureInfo.InvariantCulture );

At this point AForge.Video.DirectShow.VideoCaptureDevice class captures 
video from local camera using default settings for video size. But most 
cameras support several predefined sizes for captured video. It is 
required to have functionality, which allows to enumerate supported video 
sizes and select one of them for capturing.

Needed Color Reduction Functionality to reduce 24RGB to Indexed Palette 
with needed Colors with/wihout Dithering

What steps will reproduce the problem?
1. Open iplab
2. Click on the show histogram icon from the toolbar.
3. Now open any image
4. Close the histogram viewer.
5. Close the application.

What is the expected output? What do you see instead?
The program should exit. We get an error message say "Cannot access a
disposed object".

What version of the product are you using? On what operating system?
I am using v2.4.0 on windows XP professional. 

Please provide any additional information below.
The screen shot and the details are available in the attached file.

It is required to have a possibility to convert expression is postfix 
notation to infix

It would be nice to have support of 16 bpp grayscale images.

At this point GetObjects() method of BlobCounterBase allows to extract 
blobs only from grayscale image. The method should be extended to extract 
blobs from color images as well.