Asking about detecting the tampered/duplicated region in one image - c++

Question

Recently, I have been doing the research about how to detect the duplicated region in an image by using OpenCV+eclipse in Ubuntu.I also read the references the code of SIFT, SURF, Feature2d+Homomorphy, but these code is just the comparison between the image object and image scene? I do not know how to merge these algorithms into an image. So that I can use it to detect the duplicated region in AN image.

Answer 1

The problem is in the method through which you are setting your ROI. You are using cvSetImageROI() which is used for "IplImage" not for "Mat" images. And on the other hand, you are loading an image in Mat format. That's the reason of that error.
I have written a code. You can modify the code as per your need. The code will save the smaller images in a folder "smaller_images". So, don't forget to create a new folder with name "smaller_images".
int main( )
{
Mat image;
int width_step;
int height_step;
//image = imread( argv[1], CV_LOAD_IMAGE_COLOR);
image = imread( "myImage.jpeg", CV_LOAD_IMAGE_COLOR);
int rows = image.rows;
int cols = image.cols;
cv::Size s = image.size();
rows = s.height;
cols = s.width;
cout<<"The width of the image: "<<cols<< endl;
cout<<"The height of the image: "<<rows<< endl;
cout<< "Input your width_step= ";
cin>>width_step;
cout<< "Input your height_step= ";
cin>>height_step;
double each_width= cols/width_step;
double each_height= rows/height_step;
cout<< "The width_size of each region "<<each_width<< endl;
cout<< "The height_size of each region "<<each_height<< endl;
//-------------------------
int i=0;
for(int x=0; x<(cols - width_step); x++ )
{
for (int y=0; y < (rows - height_step ); y++)
{
if( x <(cols - width_step) && y < (rows - height_step ) )
{
Mat smallerImage;
smallerImage.create(height_step, width_step, CV_8UC3);
Rect regionOfInterest = Rect (x, y, width_step, height_step); // Rect (min_x,min_y, cols, rows);
//mySquare= original_frame(regionOfInterest);
smallerImage = image( regionOfInterest );
///Increasing the value of "x" and "y" for next image for "Rect()"
x = x + width_step;
y = y + height_step;
i = i+1;
/// Saving the smaller image to a folder called "smaller_image"
char name_writeImage[255];
sprintf(name_writeImage, "smaller_images/%d.jpg" ,i );
imwrite(name_writeImage,smallerImage );
}
}
}
waitKey(0);
return 0;
}

Answer 2

I don't know what is your application exactly but you can define some small region of interests in your image...you can also create independent images from those regions and then through a recursive loop you can check some much is the similarity between different regions.
For example: You can have a look at the following illustration where we have a bigger image and then we define few region of interests (ROI) and then we create smaller images from those ROI and then we can find similarity between those smaller images (which are actually a part of the original image)

Answer 3

you can decide yourself in how many smaller images you would like to divide your image. So, lets say that you want to create 4 smaller images from your given image whose size is 400 x 400. So, your each image will be of size 100 x 100.
Ok, now you have to create 4 images.
First image: Top left corner (0,0), width= 400/4 = 100, height = 400/4 =100
Mat image1;
image1.create(height, width, CV_8UC3);
Rect regionOfInterest_1 = Rect (0,0, width, height);
image1= original_Image(regionOfInterest_1);
Second image: Top left corner (0,101), width= 400/4 = 100, height = 400/4 =100
Mat image2;
image2.create(height, width, CV_8UC3);
Rect regionOfInterest_2 = Rect (0,101, width, height);
image2= original_Image(regionOfInterest_2);
Third image: Top left corner (0,101), width= 400/4 = 100, height = 400/4 =100
Mat image3;
image3.create(height, width, CV_8UC3);
Rect regionOfInterest_3 = Rect (101,0, width, height);
image3= original_Image(regionOfInterest_3);
and so on...you can do it using a for loop instead of writing it separately. But first you need to understand the concept of creating a smaller image from the original image using "Rect Region of interest".
One tip: Have a look at the arguments of Rect(x , y, width , height). In openCV, X-coordinate represents the columns/width and Y-coordinate represents the rows/height.