How do I get the percentage similarity of filtered objects in 2 images

Question

2.00/5 (1 vote)

See more:

I'm working on a project using AForge algorithm to filter objects i need in an image and use them in comparison to another image of the same format to see the percentage similarity of the objects. I have the code below but it only gives me the percentage similarity or difference of the whole image. I really don't care about the whole image but my interest is just the objects i filtered in the image. Please friends, how do i go about this in C# or VB.Net? Here is the code below that shows only similarity and difference of both images but not the filtered objects of interest.

VB

Private Sub Form1_Load(ByVal sender As System.Object, ByVal e As System.EventArgs) Handles MyBase.Load
    Dim bmp1 As Bitmap = Image.FromFile("C:\Temp\ImageCompare\chill.jpg")
    Dim bmp2 As Bitmap = Image.FromFile("C:\Temp\ImageCompare\chill1.jpg")

    Dim sameCount As Integer = 0
    Dim diffCount As Integer = 0

    For x As Integer = 0 To (bmp1.Width) - 1
        For y As Integer = 0 To (bmp1.Height) - 1
            If bmp1.GetPixel(x, y).Equals(bmp2.GetPixel(x, y)) Then
                sameCount += 1
            Else
                diffCount += 1
            End If
        Next
    Next

    Dim total As Integer = sameCount + diffCount

    MessageBox.Show(String.Format("Same Percentage: {0}{1}Difference Percentage: {2}", _
                                  (sameCount / total).ToString("p"), Environment.NewLine, (diffCount / total).ToString("p")))
End Sub

Posted 14-Aug-14 11:51am

Naija Coding

Add a Solution

Comments

Sergey Alexandrovich Kryukov 14-Aug-14 18:34pm

Are you able to define the concept of "similarity"? I can define a pair of images which has zero similarity in one-to-one pixel comparison and apparent visual similarity (so it would even be hard to tell the difference).
And I can define a pair of images with apparent visual difference but high similarity in pixels. Isn't it obvious that it is possible?

So, please: similarity criteria, purpose...

—SA

1 solution

Add a Solution

Add your solution here

Treat my content as plain text, not as HTML

Preview 0

…

Existing Members

Sign in to your account

...or Join us

Download, Vote, Comment, Publish.

Your Email
Password
Forgot your password?

Your Email
This email is in use. Do you need your password?
Optional Password

I have read and agree to the Terms of Service and Privacy Policy
Please subscribe me to the CodeProject newsletters

When answering a question please:

Read the question carefully.
Understand that English isn't everyone's first language so be lenient of bad spelling and grammar.
If a question is poorly phrased then either ask for clarification, ignore it, or edit the question and fix the problem. Insults are not welcome.
Don't tell someone to read the manual. Chances are they have and don't get it. Provide an answer or move on to the next question.

Let's work to help developers, not make them feel stupid.

This content, along with any associated source code and files, is licensed under The Code Project Open License (CPOL)

Sergey Alexandrovich Kryukov · Accepted Answer · 2014-08-14T12:38:00

Solution 1

Please see the comment to the question. It depends on the purpose, of course, but for most, your comparison have too little practical sense. I hope I explained why.

Now, didn't I tell you before that GetPixel is prohibitively slow? You have to use System.Drawing.Image.LockBits:
http://msdn.microsoft.com/en-us/library/system.drawing.bitmap.lockbits%28v=vs.110%29.aspx[^].

The help page on first of the two methods shows some code sample sufficient for understanding of the operation.

—SA

Posted 14-Aug-14 12:38pm

Sergey Alexandrovich Kryukov

Comments

Naija Coding 14-Aug-14 19:03pm

Ok let be clearer. The project is to design a system that automatically marks students papers in a bitmap format. This is why there are 2 images, the first is the marking sheet and the second is the student answer sheet. The idea is to use the position and colour of the objects in the marking sheet which represents the right answers to mark the objects of the students sheet.

Sergey Alexandrovich Kryukov 14-Aug-14 22:39pm

Oh, no... :-)
—SA

Naija Coding 14-Aug-14 20:21pm

Bitmap bitmap = new Bitmap(pictureBox3.Image);
BlobCounter blobCounter = new BlobCounter(bitmap);
blobCounter.MinWidth = 5;
blobCounter.MinHeight = 5;
blobCounter.FilterBlobs = true;
blobCounter.ProcessImage(bitmap);
Rectangle[] rects = blobCounter.GetObjectsRectangles();

Bitmap bitmap2 = new Bitmap(pictureBox4.Image);
BlobCounter blobCounter2 = new BlobCounter(bitmap2);
blobCounter2.MinWidth = 5;
blobCounter2.MinHeight = 5;
blobCounter2.FilterBlobs = true;
blobCounter2.ProcessImage(bitmap2);
Rectangle[] rects2 = blobCounter.GetObjectsRectangles();

foreach (Rectangle recs in rects)
foreach (Rectangle recs2 in rects2 )
if (rects.Length > 0)
{
if (rects2 .Length >0)
{
foreach (Rectangle objectRect in rects)
{
foreach (Rectangle objectRect2 in rects2 )
{

for (int x = 0; x < bitmap .Width ; x++)
{
for (int y = 0; y < bitmap.Height; y++)
{
for (int x2 = 0; x2 < bitmap2.Width; x2++)
{
for (int y2 = 0; y2 < bitmap2.Height; y2++)
{

if (bitmap .GetPixel (x,y,objectRect) == (bitmap2 .GetPixel (x2,y2,objectRect2 )))
...

this is what am trying out in c#. having errors and i dont know if am on track

Sergey Alexandrovich Kryukov 14-Aug-14 22:40pm

I just told you about GetPixel...
—SA

Naija Coding 21-Aug-14 7:06am

I have no problems with image processing, and yes the get pixel is slow and I will look into the lockbit method too. Thanks. Could it be that what I really need here is OMR aka Form processing? Because I'm already done with image processing using AForge and it works fine. What I want to do now is compare the template to the form and get the percentage of the student's result. Any help?

Sergey Alexandrovich Kryukov 21-Aug-14 10:57am

What help? Well, you need to find some OMR product or develop you own. If you think it's related to your question, it means that you approach is way too naive and hardly can be used. I think my comment to the question should explain that. Anyway, this is a different question you did not pose in first place.

As to your present question, I answered it.

—SA

Naija Coding 21-Aug-14 20:22pm

It's my first time doing this kind of project so am a bit lost, sorry about that. The omr products i checked out are costly with none less than 3000 dollars, that may not be an option as I am just a student. Do u have any other suggestions on what i should do? I somehow feel that the the location of the blobs which i have, I can do some manipulation to get this solved, I just don't know how to go about it and this is the last step of my project

Sergey Alexandrovich Kryukov 21-Aug-14 22:02pm

Then think about this: if you want to save $3000, you should be able to do the job which costs more then $3000. Isn't that logical? Which way would you rather afford? :-)
Please forgive me; I would hate to discourage you, but sometimes it's important to be: 1) logical, 2) realistic.
Believe me, I could not possibly desire more that you could do this job, but how can I help you if you are doing your best to convince me that you cannot?
For goodness sake, no offense...
—SA

Sergey Alexandrovich Kryukov 21-Aug-14 22:08pm

Now, I don't really know your settings, but perhaps your requirements are more specialized, which could make the job easier. If, for example, it's about the location of the cell which the person has to feel with the round marks (because you generate from from data, instead of scanning the form, which would be a good idea), you can really 1) use the blob recognition, 2) take the convex hull of each mark from blobs recognition; 3) using the convex hull and cell location/size, calculate degree of the match; 4) draw a conclusion that the mark feels the cell.

But if the OCR is the part of it, you should rather develop OCR, which is even less realistic, or use available OCR (is it English only, or something else?) and just integrate it with you simplified OMR.

Now, if a) your requirements are more complex, or b) you cannot implement even this simplified approach, I would dub the case hopeless. At least this is how I see it now. :-(

—SA

Naija Coding 23-Aug-14 17:36pm

I'm sorry about that. I think I can do the job, may just take a little while looking at it.

Sergey Alexandrovich Kryukov 23-Aug-14 21:44pm

Wish you the best of luck.
—SA

Naija Coding 5-Sep-14 9:20am

I have tried this but it seems very crude to me because the it has so many iterations which makes the program very slow. It works but the problem is that the program hangs because of too many iterations. Please is there something I'm missing? I'm I on the right track or something? Please help and pardon my crude codes below.
Bitmap bitmap = new Bitmap(pictureBox1.Image);
Bitmap bitmap2 = new Bitmap(pictureBox2.Image);

// create an instance of blob counter algorithm
BlobCounter blobCounter = new BlobCounter();
blobCounter.MinWidth = 5;
blobCounter.MinHeight = 5;
blobCounter.FilterBlobs = true;
blobCounter.ObjectsOrder = ObjectsOrder.Size;
blobCounter.ProcessImage(bitmap);
BlobCounter blobCounter2 = new BlobCounter();
blobCounter2.MinWidth = 5;
blobCounter2.MinHeight = 5;
blobCounter2.FilterBlobs = true;
blobCounter2.ObjectsOrder = ObjectsOrder.Size;
blobCounter2.ProcessImage(bitmap2);

Rectangle[] rects = blobCounter.GetObjectsRectangles();

Rectangle[] rects2 = blobCounter2.GetObjectsRectangles();

foreach (Rectangle recs in rects)
foreach (Rectangle recs2 in rects2)

if (rects.Length > 0 )
{
if (rects2.Length > 0)
{
for (int x = 0; x < recs.Width & x < recs.Height; x++)
{
// for (int x2 = 0; x2 < recs2.Width; x2++)
for (int y = 0; y < recs2.Width & y < recs2.Height; y++)
{
// for (int y2 = 0; y2 < recs2.Height; y2++)

if (recs.Equals(recs2))
{
this.Refresh();
//listBox1.Items.Add("TRUE");
Console.WriteLine("TRUE");

}
else
{
//listBox1.Items.Add("FALSE");
Console.WriteLine("FALSE");
}
.....

Sergey Alexandrovich Kryukov 5-Sep-14 10:11am

This is slow because your algorithm is of O(N²) complexity, where N: the number if the blobs on one image. Could you try linear comparison? What if you assume that, if images have the same origins and slightly different, assume that blob analysis put the blob in the same order? And then you would compare first blob with first, second with second, and so one. It will give you much faster, O(N) complexity. Not very apparent, and less of the comparison "validity" but might work for your purpose. Would you try?
—SA

Naija Coding 7-Sep-14 18:48pm

I solved it. The slowness was because it was iterating to the width and height of the rectangular objects. i just made it to iterate once and it works well. I think I'm done with my project, I just hope my supervisor will be pleased when I see him next week. THANKS FOR YOUR HELP.

Sergey Alexandrovich Kryukov 7-Sep-14 21:10pm

Great!
—SA

Naija Coding 26-Sep-14 19:56pm

Thank u. My supervisor has approved it. Thanks once again for been there. I'm now in the report stage of the project. program is working well and fine.

Sergey Alexandrovich Kryukov 26-Sep-14 20:34pm

Do you know the pleasure of helping someone who actually can get advice and make things working?
Thank you for sending me good news.
Good luck, call again.
—SA

Naija Coding 18-Dec-14 14:46pm

Hi. It's been a long time. My software works very well if computer is used to fill the blobs but if they are filled by humans, it doesn't work well because the location of some blobs will change if humans fill it. how do i resolve this? is there anything i can do? thanks

Sergey Alexandrovich Kryukov 18-Dec-14 14:51pm

I don't know. I would separate what humans can do and what a machine does. Doesn't it look more like organizational anti-pattern problem?
—SA

Naija Coding 19-Dec-14 7:11am

problem is I'm submitting the work finally in 2 weeks and i dont know what to do :-(

Naija Coding 18-Dec-14 15:12pm

seems like. i used template matching algorithm so it can compare what the teacher fills with what the students fills. obviously as humans, the student will fill the blobs slightly different from the teacher. that slight difference in most cases returns a non equal value whereas it is the same option selected by the student. i tried inflating the blobs and filling it but it still remains same because the blob is filled and inflated with respect to the original fill.