How to detect the edge of the photo

I want to detect the photo edge and I turn the outside black color. However the thing is there are some dirty dots on the photo. So when I detect the edge, the dirty pot become the darkest and there is a line appear. I want to know how to remove it.

Below is my program can someone help me check what should I do. Thanks a lot.

I1=imread('C:\Users\Lenovo\Desktop\FYP_P055\Huijie1.jpg');
A=imread('C:\Users\Lenovo\Desktop\FYP_P055\Huijie1.jpg');

imtool(A);

hist(double(A));

I1(I1<40)=0;

A1=hist(double(I1));

hist(double(I1));

[row,col]=size(I1);

I11=I1(1:row/2,1:col/2);

I12=I1(1:row/2,col/2:end);

I21=I1(row/2+1:end,1:col/2);

I22=I1(row/2+1:end,col/2:end);

M11=min(I11,[],1);

M12=min(I12,[],1);

M21=min(I21,[],1);

M22=min(I22,[],1);

for c=1:col/2

      for r=1:row/2

          if I1(r,c)==M11(c)

              min_val_col=c;

              min_val_row=r;

              for cc=1:min_val_row

                  I1(cc,min_val_col)=0;

              end
          end
      end

     for r=row/2+1:row
           if I1(r,c)==M21(c)
               min_val_col=c;
               min_val_row=r;
                   for cc=min_val_row:row
                       I1(cc,min_val_col)=0;
                   end
           end
       end
  end
   for c2=round(col/2):col
       c=c2-round(col/2)+1;
       for r=1:row/2
           if I1(r,c2)==M12(c)
               min_val_col=c2;
               min_val_row=r;
               for cc=1:min_val_row
                   I1(cc,min_val_col)=0;
               end
           end
       end
       for r=row/2+1:row
           if I1(r,c2)==M22(c)
               min_val_col=c2;
               min_val_row=r;
                   for cc=min_val_row:row
                       I1(cc,min_val_col)=0;
                   end
           end
       end
   end
  imtool(I1);

[Sivakumaran Chandrasekaran]

check demos of image analyst in this website. you will get

[Walter Roberson]

imclearborder() ?

Answer 1

[Image Analyst]

Easy. Threshold. Take the largest blob, then get its convex hull. See attached demo script below this image it creates:

[huijie Q]

I still cannot get it. Should I delete my program first or need to add your program in?

[huijie Q]

I cannot get the final masked and the masked image.

[Image Analyst]

Why not? You can turn my code into a function called GetMaskedImage() or something. What went wrong?

[huijie Q]

Thank you for answering me. But I want it to be detect column by column see which part has lesser intensity. Then It will turn black. If the point is more far away from the original edge, then I reject that point and choose the point near the edge.

[Image Analyst]

I read it several times but I just cannot understand what that means. Please attach a final output image so I can see what you want.

[huijie Q]

I can get the first 3 already. Can you show me your program for how to get final mask and masked image? thanks!

[Image Analyst]

I still don't understand. The demo gives all the images in the figure, including the final mask and masked image. Why do you think it doesn't?

Answer 2

[Thorsten]

You can use imfindcircles to detect the circle, and then set all points outside the circle to black. See http://stackoverflow.com/questions/29323640/circle-detection-via-hough-transform