We Keep Coding

OpenCV: Getting x co-ordinate from rect.br()

I am able to detect and draw a bounding rectangle around a region of interest like so:
Now I need to find the "x" value of the bottom horizontal line of the green bounding rectangle. My main aim is to find the "x" value of the base of the battery, and the "x" value of the blue strip of paper, so that I can calculate the distance between them.
There are only 2 values, rect.tl() and rect.br(), that I can see be used to draw the bounding rectangle Imgproc.rectangle(sourceMat, rect.tl(), rect.br(), green, 3);. I'm under the assumption that the "x" value from the bottom right point of the bounding rectangle will be the x co-ordinate of the pixel point.
How do i get the 'x' value from rect.br()
When i print out rect.br() like so Log.e("BR", rect.br().toString()); i get this:
E/BR: {793.0, 1686.0}
private Bitmap findRoi(Bitmap sourceBitmap) {
Bitmap roiBitmap = null;
Scalar green = new Scalar(0, 255, 0, 255);
Mat sourceMat = new Mat(sourceBitmap.getWidth(), sourceBitmap.getHeight(), CvType.CV_8UC3);
Utils.bitmapToMat(sourceBitmap, sourceMat);
Mat roiTmp = sourceMat.clone();
final Mat hsvMat = new Mat();
sourceMat.copyTo(hsvMat);
// convert mat to HSV format for Core.inRange()
Imgproc.cvtColor(hsvMat, hsvMat, Imgproc.COLOR_RGB2HSV);
Scalar lowerb = new Scalar(85, 50, 40); // lower color border for BLUE
Scalar upperb = new Scalar(135, 255, 255); // upper color border for BLUE
Core.inRange(hsvMat, lowerb, upperb, roiTmp); // select only blue pixels
// find contours
List<MatOfPoint> contours = new ArrayList<>();
List<RotatedRect> boundingRects = new ArrayList<>();
Imgproc.findContours(roiTmp, contours, new Mat(), Imgproc.RETR_LIST, Imgproc.CHAIN_APPROX_SIMPLE);
// find appropriate bounding rectangles
for (MatOfPoint contour : contours) {
MatOfPoint2f areaPoints = new MatOfPoint2f(contour.toArray());
RotatedRect boundingRect = Imgproc.minAreaRect(areaPoints);
double rectangleArea = boundingRect.size.area();
// test min ROI area in pixels
if (rectangleArea > 40001) {//400000
Point rotated_rect_points[] = new Point[4];
boundingRect.points(rotated_rect_points);
Rect rect = Imgproc.boundingRect(new MatOfPoint(rotated_rect_points));
// test horizontal ROI orientation
if (rect.width > rect.height) {
Imgproc.rectangle(sourceMat, rect.tl(), rect.br(), green, 3);
}
}
}
roiBitmap = Bitmap.createBitmap(sourceMat.cols(), sourceMat.rows(), Bitmap.Config.ARGB_8888);
Utils.matToBitmap(sourceMat, roiBitmap);
return roiBitmap;
}

I don't know a lot about Java but AFAIK it must be similar to c++, so whether you know about it or not,I try my code in c++. Suppose that you are working on the below rectangle:
For having the above Rect in OpenCV:
Rect r = Rect(20,20,30,20); // Rect(x,y,width,height)
And we could access the below-right x with using the br() like the below.
int x2=r.br().x ;
Below pic will tell you the whole story.

Best parameters for pupil detection using hough? java opencv

--------------read edit below---------------
I am trying to detect the edge of the pupils and iris within various images. I am altering parameters and such but I can only manage to ever get one iris/pupil outline correct, or get unnecessary outlines in the background, or none at all. Is the some specific parameters that I should try to try and get the correct outlines. Or is there a way that I can crop the image just to the eyes, so the system can focus on that part?
This is my UPDATED method:
private void findPupilIris() throws IOException {
//converts and saves image in grayscale
Mat newimg = Imgcodecs.imread("/Users/.../pic.jpg");
Mat des = new Mat(newimg.rows(), newimg.cols(), newimg.type());
Mat norm = new Mat();
Imgproc.cvtColor(newimg, des, Imgproc.COLOR_BGR2HSV);
List<Mat> hsv = new ArrayList<Mat>();
Core.split(des, hsv);
Mat v = hsv.get(2); //gets the grey scale version
Imgcodecs.imwrite("/Users/Lisa-Maria/Documents/CapturedImages/B&Wpic.jpg", v); //only writes mats
CLAHE clahe = Imgproc.createCLAHE(2.0, new Size(8,8) ); //2.0, new Size(8,8)
clahe.apply(v,v);
// Imgproc.GaussianBlur(v, v, new Size(9,9), 3); //adds left pupil boundary and random circle on 'a'
// Imgproc.GaussianBlur(v, v, new Size(9,9), 13); //adds right outer iris boundary and random circle on 'a'
Imgproc.GaussianBlur(v, v, new Size(9,9), 7); //adds left outer iris boundary and random circle on left by hair
// Imgproc.GaussianBlur(v, v, new Size(7,7), 15);
Core.addWeighted(v, 1.5, v, -0.5, 0, v);
Imgcodecs.imwrite("/Users/.../after.jpg", v); //only writes mats
if (v != null) {
Mat circles = new Mat();
Imgproc.HoughCircles( v, circles, Imgproc.CV_HOUGH_GRADIENT, 2, v.rows(), 100, 20, 20, 200 );
List<MatOfPoint> contours = new ArrayList<MatOfPoint>();
System.out.println("circles.cols() " + circles.cols());
if(circles.cols() > 0) {
System.out.println("1");
for (int x = 0; x < circles.cols(); x++) {
System.out.println("2");
double vCircle[] = circles.get(0, x);
if(vCircle == null) {
break;
}
Point pt = new Point(Math.round(vCircle[0]), Math.round(vCircle[1]));
int radius = (int) Math.round(vCircle[2]);
//draw the found circle
Imgproc.circle(v, pt, radius, new Scalar(255,0,0),2); //newimg
//Imgproc.circle(des, pt, radius/3, new Scalar(225,0,0),2); //pupil
Imgcodecs.imwrite("/Users/.../Houghpic.jpg", v); //newimg
//draw the mask: white circle on black background
// Mat mask = new Mat( new Size( des.cols(), des.rows() ), CvType.CV_8UC1 );
// Imgproc.circle(mask, pt, radius, new Scalar(255,0,0),2);
// des.copyTo(des,mask);
// Imgcodecs.imwrite("/Users/..../mask.jpg", des); //newimg
Imgproc.logPolar(des, norm, pt, radius, Imgproc.WARP_FILL_OUTLIERS);
Imgcodecs.imwrite("/Users/..../Normalised.jpg",norm);
}
}
}
}
Result: hough pic

Following discussion in comments, I am posting a general answer with some results I got on the worst case image uploaded by the OP.
Note : The code I am posting is in Python, since it is the fastest for me to write
Step 1. As you ask for a way to crop the image, so as to focus on the eyes only, you might want to look at Face Detection. Since, the image essentially requires to find eyes only, I did the following:
eye_cascade = cv2.CascadeClassifier('haarcascade_eye.xml')
eyes = eye_cascade.detectMultiScale(v) // v is the value channel of the HSV image
// The results "eyes" gives you the dimensions of the rectangle where the eyes are detected as [x, y, w, h]
// Just for drawing
cv2.rectangle(v, (x1, y1), (x1+w1, y1+h1), (0, 255, 0), 2)
cv2.rectangle(v, (x2, y2), (x2+w2, y2+h2), (0, 255, 0), 2)
Now, once you have the bounding rectangles, you can crop the rectangles from the image like:
crop_eye1 = v[y1:y1+h1, x1:x1+w1]
crop_eye2 = v[y2:y2+h2, x2:x2+w2]
After you obtain the rectangles, I would suggest looking into different color spaces instead of RGB/BGR, HSV/Lab/Luv in particular.
Because the R, G, and B components of an object’s color in a digital image are all correlated with the amount of light hitting the object, and therefore with each other, image descriptions in terms of those components make object discrimination difficult. Descriptions in terms of hue/lightness/chroma or hue/lightness/saturation are often more relevant
Then, once, you have the eyes, its time to equalize the contrast of the image, however, I suggest using CLAHE and play with the parameters for clipLimit and tileGridSize. Here is a code which I implemented a while back in Java:
private static Mat clahe(Mat image, int ClipLimit, Size size){
CLAHE clahe = Imgproc.createCLAHE();
clahe.setClipLimit(ClipLimit);
clahe.setTilesGridSize(size);
Mat dest_image = new Mat();
clahe.apply(image, dest_image);
return dest_image;
}
Once you are satisfied, you should sharpen the image so that HoughCircle is robust. You should look at unsharpMask. Here is the code in Java for UnsharpMask I implemented in Java:
private static Mat unsharpMask(Mat input_image, Size size, double sigma){
// Make sure the {input_image} is gray.
Mat sharpend_image = new Mat(input_image.rows(), input_image.cols(), input_image.type());
Mat Blurred_image = new Mat(input_image.rows(), input_image.cols(), input_image.type());
Imgproc.GaussianBlur(input_image, Blurred_image, size, sigma);
Core.addWeighted(input_image, 2.0D, Blurred_image, -1.0D, 0.0D, sharpened_image);
return sharpened_image;
}
Alternatively, you could use bilateral filter, which is edge preserving smoothing, or read through this for defining a custom kernel for sharpening image.
Hope it helps and best of luck!

Crop out part from images (findContours?) {opencv, java}

I have this image with boxes containing letters, like this:
I have been able to crop out each box., like this:
Now to my question. How can i crop out the letters only from each box? The desired result looks like this
I would like to use findContours but I am not really sure of how to achieve this since it will detect the noise and everything around as well.

Approach
I suggest the following approach according to this fact that you can extract the box. If you are give the box follow the steps, I think that would work:
Find the center of image
Find the contours in the image - those can be candidates
Find the bounding rectangle of each contour
Find the center of each bounding rectangle
Find the distance of each bounding rectangle from the center of image
Find the minimum distance - your answer
Note: There is a var named pad which control the padding of the result figure!
Do this for all your boxes. I hope that will help!
Good Luck :)
Python Code
# reading image in grayscale
image = cv2.imread('testing2.jpg',cv2.CV_LOAD_IMAGE_GRAYSCALE)
# thresholding to get a binary one
ret, image = cv2.threshold(image, 100,255,cv2.THRESH_BINARY_INV)
# finding the center of image
image_center = (image.shape[0]/2, image.shape[1]/2)
if image is None:
print 'can not read the image data'
# finding image contours
contours, hier = cv2.findContours(image, cv2.RETR_EXTERNAL, cv2.CHAIN_APPROX_SIMPLE)
# finding distance of each contour from the center of image
d_min = 1000
for contour in contours:
# finding bounding rect
rect = cv2.boundingRect(contour)
# skipping the outliers
if rect[3] > image.shape[1]/2 and rect[2] > image.shape[0]/2:
continue
pt1 = (rect[0], rect[1])
# finding the center of bounding rect-digit
c = (rect[0]+rect[2]*1/2, rect[1]+rect[3]*1/2)
d = np.sqrt((c[0] - image_center[0])**2 + (c[1]-image_center[1])**2)
# finding the minimum distance from the center
if d < d_min:
d_min = d
rect_min = [pt1, (rect[2],rect[3])]
# fetching the image with desired padding
pad = 5
result = image[rect_min[0][1]-pad:rect_min[0][1]+rect_min[1][1]+pad, rect_min[0][0]-pad:rect_min[0][0]+rect_min[1][0]+pad]
plt.imshow(result*255, 'gray')
plt.show()
Java Code
System.loadLibrary(Core.NATIVE_LIBRARY_NAME);
// reading image
Mat image = Highgui.imread(".\\testing2.jpg", Highgui.CV_LOAD_IMAGE_GRAYSCALE);
// clone the image
Mat original = image.clone();
// thresholding the image to make a binary image
Imgproc.threshold(image, image, 100, 128, Imgproc.THRESH_BINARY_INV);
// find the center of the image
double[] centers = {(double)image.width()/2, (double)image.height()/2};
Point image_center = new Point(centers);
// finding the contours
ArrayList<MatOfPoint> contours = new ArrayList<MatOfPoint>();
Mat hierarchy = new Mat();
Imgproc.findContours(image, contours, hierarchy, Imgproc.RETR_EXTERNAL, Imgproc.CHAIN_APPROX_SIMPLE);
// finding best bounding rectangle for a contour whose distance is closer to the image center that other ones
double d_min = Double.MAX_VALUE;
Rect rect_min = new Rect();
for (MatOfPoint contour : contours) {
Rect rec = Imgproc.boundingRect(contour);
// find the best candidates
if (rec.height > image.height()/2 & rec.width > image.width()/2)
continue;
Point pt1 = new Point((double)rec.x, (double)rec.y);
Point center = new Point(rec.x+(double)(rec.width)/2, rec.y + (double)(rec.height)/2);
double d = Math.sqrt(Math.pow((double)(pt1.x-image_center.x),2) + Math.pow((double)(pt1.y -image_center.y), 2));
if (d < d_min)
{
d_min = d;
rect_min = rec;
}
}
// slicing the image for result region
int pad = 5;
rect_min.x = rect_min.x - pad;
rect_min.y = rect_min.y - pad;
rect_min.width = rect_min.width + 2*pad;
rect_min.height = rect_min.height + 2*pad;
Mat result = original.submat(rect_min);
Highgui.imwrite("result.png", result);
EDIT:
Java code added!
Result

Stitching 2 images (OpenCV)

I'm trying to stitch two images together, using the OpenCV Java API. However, I get the wrong output and I cannot work out the problem. I use the following steps:
1. detect features
2. extract features
3. match features.
4. find homography
5. find perspective transform
6. warp perspective
7. 'stitch' the 2 images, into a combined image.
but somewhere I'm going wrong. I think it's the way I'm combing the 2 images, but I'm not sure. I get 214 good feature matches between the 2 images, but cannot stitch them?
public class ImageStitching {
static Mat image1;
static Mat image2;
static FeatureDetector fd;
static DescriptorExtractor fe;
static DescriptorMatcher fm;
public static void initialise(){
fd = FeatureDetector.create(FeatureDetector.BRISK);
fe = DescriptorExtractor.create(DescriptorExtractor.SURF);
fm = DescriptorMatcher.create(DescriptorMatcher.BRUTEFORCE);
//images
image1 = Highgui.imread("room2.jpg");
image2 = Highgui.imread("room3.jpg");
//structures for the keypoints from the 2 images
MatOfKeyPoint keypoints1 = new MatOfKeyPoint();
MatOfKeyPoint keypoints2 = new MatOfKeyPoint();
//structures for the computed descriptors
Mat descriptors1 = new Mat();
Mat descriptors2 = new Mat();
//structure for the matches
MatOfDMatch matches = new MatOfDMatch();
//getting the keypoints
fd.detect(image1, keypoints1);
fd.detect(image1, keypoints2);
//getting the descriptors from the keypoints
fe.compute(image1, keypoints1, descriptors1);
fe.compute(image2,keypoints2,descriptors2);
//getting the matches the 2 sets of descriptors
fm.match(descriptors2,descriptors1, matches);
//turn the matches to a list
List<DMatch> matchesList = matches.toList();
Double maxDist = 0.0; //keep track of max distance from the matches
Double minDist = 100.0; //keep track of min distance from the matches
//calculate max & min distances between keypoints
for(int i=0; i<keypoints1.rows();i++){
Double dist = (double) matchesList.get(i).distance;
if (dist<minDist) minDist = dist;
if(dist>maxDist) maxDist=dist;
}
System.out.println("max dist: " + maxDist );
System.out.println("min dist: " + minDist);
//structure for the good matches
LinkedList<DMatch> goodMatches = new LinkedList<DMatch>();
//use only the good matches (i.e. whose distance is less than 3*min_dist)
for(int i=0;i<descriptors1.rows();i++){
if(matchesList.get(i).distance<3*minDist){
goodMatches.addLast(matchesList.get(i));
}
}
//structures to hold points of the good matches (coordinates)
LinkedList<Point> objList = new LinkedList<Point>(); // image1
LinkedList<Point> sceneList = new LinkedList<Point>(); //image 2
List<KeyPoint> keypoints_objectList = keypoints1.toList();
List<KeyPoint> keypoints_sceneList = keypoints2.toList();
//putting the points of the good matches into above structures
for(int i = 0; i<goodMatches.size(); i++){
objList.addLast(keypoints_objectList.get(goodMatches.get(i).queryIdx).pt);
sceneList.addLast(keypoints_sceneList.get(goodMatches.get(i).trainIdx).pt);
}
System.out.println("\nNum. of good matches" +goodMatches.size());
MatOfDMatch gm = new MatOfDMatch();
gm.fromList(goodMatches);
//converting the points into the appropriate data structure
MatOfPoint2f obj = new MatOfPoint2f();
obj.fromList(objList);
MatOfPoint2f scene = new MatOfPoint2f();
scene.fromList(sceneList);
//finding the homography matrix
Mat H = Calib3d.findHomography(obj, scene);
//LinkedList<Point> cornerList = new LinkedList<Point>();
Mat obj_corners = new Mat(4,1,CvType.CV_32FC2);
Mat scene_corners = new Mat(4,1,CvType.CV_32FC2);
obj_corners.put(0,0, new double[]{0,0});
obj_corners.put(0,0, new double[]{image1.cols(),0});
obj_corners.put(0,0,new double[]{image1.cols(),image1.rows()});
obj_corners.put(0,0,new double[]{0,image1.rows()});
Core.perspectiveTransform(obj_corners, scene_corners, H);
//structure to hold the result of the homography matrix
Mat result = new Mat();
//size of the new image - i.e. image 1 + image 2
Size s = new Size(image1.cols()+image2.cols(),image1.rows());
//using the homography matrix to warp the two images
Imgproc.warpPerspective(image1, result, H, s);
int i = image1.cols();
Mat m = new Mat(result,new Rect(i,0,image2.cols(), image2.rows()));
image2.copyTo(m);
Mat img_mat = new Mat();
Features2d.drawMatches(image1, keypoints1, image2, keypoints2, gm, img_mat, new Scalar(254,0,0),new Scalar(254,0,0) , new MatOfByte(), 2);
//creating the output file
boolean imageStitched = Highgui.imwrite("imageStitched.jpg",result);
boolean imageMatched = Highgui.imwrite("imageMatched.jpg",img_mat);
}
public static void main(String args[]){
System.loadLibrary(Core.NATIVE_LIBRARY_NAME);
initialise();
}
I cannot embed images nor post more than 2 links, because of reputation points? so I've linked the incorrectly stitched images and an image showing the matched features between the 2 images (to get an understanding of the issue):
incorrect stitched image: http://oi61.tinypic.com/11ac01c.jpg
detected features: http://oi57.tinypic.com/29m3wif.jpg

It seems that you have a lot of outliers that make the estimation of homography is incorrect. SO you can use RANSAC method that recursively reject those outliers.
No need much efforts for that, just use a third parameter in findHomography function as:
Mat H = Calib3d.findHomography(obj, scene, CV_RANSAC);
Edit
Then try to be sure that your images given to detector are 8-bit grayscale image, as mentioned here

The "incorrectly stitched image" you post looks like having a bad conditioned H matrix. Apart from +dervish suggestions, run:
cv::determinant(H) > 0.01
To check if your H matrix is "usable". If the matrix is badly conditioned, you get the effect you are showing.
You are drawing onto a 2x2 canvas size, if that's the case, you won't see plenty of stitching configurations, i.e. it's ok for image A on the left of image B but not otherwise. Try drawing the output onto a 3x3 canvas size, using the following snippet:
// Use the Homography Matrix to warp the images, but offset it to the
// center of the output canvas. Careful to pre-multiply, not post-multiply.
cv::Mat Offset = (cv::Mat_<double>(3,3) << 1, 0,
width, 0, 1, height, 0, 0, 1);
H = Offset * H;
cv::Mat result;
cv::warpPerspective(mat_l,
result,
H,
cv::Size(3*width, 3*height));
// Copy the reference image to the center of the 3x3 output canvas.
cv::Mat roi = result.colRange(width,2*width).rowRange(height,2*height);
mat_r.copyTo(roi);
Where width and height are those of the input images, supposedly both of the same size. Note that this warping assumes the mat_l unchanged (flat) and mat_r warping to get stitched on it.

Image Processing Edge Detection in Java

This is my situation. It involves aligning a scanned image which will account for incorrect scanning. I must align the scanned image with my Java program.
These are more details:
There is a table-like form printed on a sheet of paper, which will be scanned into an image file.
I will open the picture with Java, and I will have an OVERLAY of text boxes.
The text boxes are supposed to align correctly with the scanned image.
In order to align correctly, my Java program must analyze the scanned image and detect the coordinates of the edges of the table on the scanned image, and thus position the image and the textboxes so that the textboxes and the image both align properly (in case of incorrect scanning)
You see, the guy scanning the image might not necessarily place the image in a perfectly correct position, so I need my program to automatically align the scanned image as it loads it. This program will be reusable on many of such scanned images, so I need the program to be flexible in this way.
My question is one of the following:
How can I use Java to detect the y coordinate of the upper edge of the table and the x-coordinate of the leftmost edge of the table. The table is a a regular table with many cells, with black thin border, printed on a white sheet of paper (horizontal printout)
If an easier method exists to automatically align the scanned image in such a way that all scanned images will have the graphical table align to the same x, y coordinates, then share this method :).
If you don't know the answer to the above to questions, do tell me where I should start. I don't know much about graphics java programming and I have about 1 month to finish this program. Just assume that I have a tight schedule and I have to make the graphics part as simple as possible for me.
Cheers and thank you.

Try to start from a simple scenario and then improve the approach.
Detect corners.
Find the corners in the boundaries of the form.
Using the form corners coordinates, calculate the rotation angle.
Rotate/scale the image.
Map the position of each field in the form relative to form origin coordinates.
Match the textboxes.
The program presented at the end of this post does the steps 1 to 3. It was implemented using Marvin Framework. The image below shows the output image with the detected corners.
The program also outputs: Rotation angle:1.6365770416167182
Source code:
import java.awt.Color;
import java.awt.Point;
import marvin.image.MarvinImage;
import marvin.io.MarvinImageIO;
import marvin.plugin.MarvinImagePlugin;
import marvin.util.MarvinAttributes;
import marvin.util.MarvinPluginLoader;
public class FormCorners {
public FormCorners(){
// Load plug-in
MarvinImagePlugin moravec = MarvinPluginLoader.loadImagePlugin("org.marvinproject.image.corner.moravec");
MarvinAttributes attr = new MarvinAttributes();
// Load image
MarvinImage image = MarvinImageIO.loadImage("./res/printedForm.jpg");
// Process and save output image
moravec.setAttribute("threshold", 2000);
moravec.process(image, null, attr);
Point[] boundaries = boundaries(attr);
image = showCorners(image, boundaries, 12);
MarvinImageIO.saveImage(image, "./res/printedForm_output.jpg");
// Print rotation angle
double angle = (Math.atan2((boundaries[1].y*-1)-(boundaries[0].y*-1),boundaries[1].x-boundaries[0].x) * 180 / Math.PI);
angle = angle >= 0 ? angle : angle + 360;
System.out.println("Rotation angle:"+angle);
}
private Point[] boundaries(MarvinAttributes attr){
Point upLeft = new Point(-1,-1);
Point upRight = new Point(-1,-1);
Point bottomLeft = new Point(-1,-1);
Point bottomRight = new Point(-1,-1);
double ulDistance=9999,blDistance=9999,urDistance=9999,brDistance=9999;
double tempDistance=-1;
int[][] cornernessMap = (int[][]) attr.get("cornernessMap");
for(int x=0; x<cornernessMap.length; x++){
for(int y=0; y<cornernessMap[0].length; y++){
if(cornernessMap[x][y] > 0){
if((tempDistance = Point.distance(x, y, 0, 0)) < ulDistance){
upLeft.x = x; upLeft.y = y;
ulDistance = tempDistance;
}
if((tempDistance = Point.distance(x, y, cornernessMap.length, 0)) < urDistance){
upRight.x = x; upRight.y = y;
urDistance = tempDistance;
}
if((tempDistance = Point.distance(x, y, 0, cornernessMap[0].length)) < blDistance){
bottomLeft.x = x; bottomLeft.y = y;
blDistance = tempDistance;
}
if((tempDistance = Point.distance(x, y, cornernessMap.length, cornernessMap[0].length)) < brDistance){
bottomRight.x = x; bottomRight.y = y;
brDistance = tempDistance;
}
}
}
}
return new Point[]{upLeft, upRight, bottomRight, bottomLeft};
}
private MarvinImage showCorners(MarvinImage image, Point[] points, int rectSize){
MarvinImage ret = image.clone();
for(Point p:points){
ret.fillRect(p.x-(rectSize/2), p.y-(rectSize/2), rectSize, rectSize, Color.red);
}
return ret;
}
public static void main(String[] args) {
new FormCorners();
}
}

Edge detection is something that is typically done by enhancing the contrast between neighboring pixels, such that you get a easily detectable line, which is suitable for further processing.
To do this, a "kernel" transforms a pixel according it the pixel's inital value, and the value of that pixel's neighbors. A good edge detection kernel will enhance the differences between neighboring pixels, and reduce the strength of a pixel with similar neigbors.
I would start by looking at the Sobel operator. This might not return results that are immediately useful to you; however, it will get you far closer than you would be if you were to approach the problem with little knowledge of the field.
After you have some crisp clean edges, you can use larger kernels to detect points where it seems that a 90% bend in two lines occurs, that might give you the pixel coordinates of the outer rectangle, which might be enough for your purposes.
With those outer coordinates, it still is a bit of math to make the new pixels be composted with the average values between the old pixels rotated and moved to "match". The results (especially if you do not know about anti-aliasing math) can be pretty bad, adding blur to the image.
Sharpening filters might be a solution, but they come with their own issues, mainly they make the picture sharper by adding graininess. Too much, and it is obvious that the original image is not a high-quality scan.

I researched the libraries but in the end I found it more convenient to code up my own edge detection methods.
The class below will detect black/grayed out edges of a scanned sheet of paper that contains such edges, and will return the x and y coordinate of the edges of the sheet of paper, starting from the rightmost end (reverse = true) or from lower end (reverse = true) or from the top edge (reverse = false) or from left edge (reverse = false). Also...the program will take ranges along vertical edges (rangex) measured in pixels, and horizontal ranges (rangey) measured in pixels. The ranges determine outliers in the points received.
The program does 4 vertical cuts using the specified arrays, and 4 horizontal cuts. It retrieves the values of the dark dots. It uses the ranges to eliminate outliers. Sometimes, a little spot on the paper may cause an outlier point. The smaller the range, the fewer the outliers. However, sometimes the edge is slightly tilted, so you don't want to make the range too small.
Have fun. It works perfectly for me.
import java.awt.image.BufferedImage;
import java.awt.Color;
import java.util.ArrayList;
import java.lang.Math;
import java.awt.Point;
public class EdgeDetection {
public App ap;
public int[] horizontalCuts = {120, 220, 320, 420};
public int[] verticalCuts = {300, 350, 375, 400};
public void printEdgesTest(BufferedImage image, boolean reversex, boolean reversey, int rangex, int rangey){
int[] mx = horizontalCuts;
int[] my = verticalCuts;
//you are getting edge points here
//the "true" parameter indicates that it performs a cut starting at 0. (left edge)
int[] xEdges = getEdges(image, mx, reversex, true);
int edgex = getEdge(xEdges, rangex);
for(int x = 0; x < xEdges.length; x++){
System.out.println("EDGE = " + xEdges[x]);
}
System.out.println("THE EDGE = " + edgex);
//the "false" parameter indicates you are doing your cut starting at the end (image.getHeight)
//and ending at 0
//if the parameter was true, it would mean it would start the cuts at y = 0
int[] yEdges = getEdges(image, my, reversey, false);
int edgey = getEdge(yEdges, rangey);
for(int y = 0; y < yEdges.length; y++){
System.out.println("EDGE = " + yEdges[y]);
}
System.out.println("THE EDGE = " + edgey);
}
//This function takes an array of coordinates...detects outliers,
//and computes the average of non-outlier points.
public int getEdge(int[] edges, int range){
ArrayList<Integer> result = new ArrayList<Integer>();
boolean[] passes = new boolean[edges.length];
int[][] differences = new int[edges.length][edges.length-1];
//THIS CODE SEGMENT SAVES THE DIFFERENCES BETWEEN THE POINTS INTO AN ARRAY
for(int n = 0; n<edges.length; n++){
for(int m = 0; m<edges.length; m++){
if(m < n){
differences[n][m] = edges[n] - edges[m];
}else if(m > n){
differences[n][m-1] = edges[n] - edges[m];
}
}
}
//This array determines which points are outliers or nots (fall within range of other points)
for(int n = 0; n<edges.length; n++){
passes[n] = false;
for(int m = 0; m<edges.length-1; m++){
if(Math.abs(differences[n][m]) < range){
passes[n] = true;
System.out.println("EDGECHECK = TRUE" + n);
break;
}
}
}
//Create a new array only using valid points
for(int i = 0; i<edges.length; i++){
if(passes[i]){
result.add(edges[i]);
}
}
//Calculate the rounded mean... This will be the x/y coordinate of the edge
//Whether they are x or y values depends on the "reverse" variable used to calculate the edges array
int divisor = result.size();
int addend = 0;
double mean = 0;
for(Integer i : result){
addend += i;
}
mean = (double)addend/(double)divisor;
//returns the mean of the valid points: this is the x or y coordinate of your calculated edge.
if(mean - (int)mean >= .5){
System.out.println("MEAN " + mean);
return (int)mean+1;
}else{
System.out.println("MEAN " + mean);
return (int)mean;
}
}
//this function computes "dark" points, which include light gray, to detect edges.
//reverse - when true, starts counting from x = 0 or y = 0, and ends at image.getWidth or image.getHeight()
//verticalEdge - determines whether you want to detect a vertical edge, or a horizontal edge
//arr[] - determines the coordinates of the vertical or horizontal cuts you will do
//set the arr[] array according to the graphical layout of your scanned image
//image - this is the image you want to detect black/white edges of
public int[] getEdges(BufferedImage image, int[] arr, boolean reverse, boolean verticalEdge){
int red = 255;
int green = 255;
int blue = 255;
int[] result = new int[arr.length];
for(int n = 0; n<arr.length; n++){
for(int m = reverse ? (verticalEdge ? image.getWidth():image.getHeight())-1:0; reverse ? m>=0:m<(verticalEdge ? image.getWidth():image.getHeight());){
Color c = new Color(image.getRGB(verticalEdge ? m:arr[n], verticalEdge ? arr[n]:m));
red = c.getRed();
green = c.getGreen();
blue = c.getBlue();
//determine if the point is considered "dark" or not.
//modify the range if you want to only include really dark spots.
//occasionally, though, the edge might be blurred out, and light gray helps
if(red<239 && green<239 && blue<239){
result[n] = m;
break;
}
//count forwards or backwards depending on reverse variable
if(reverse){
m--;
}else{
m++;
}
}
}
return result;
}
}

A similar such problem I've done in the past basically figured out the orientation of the form, re-aligned it, re-scaled it, and I was all set. You can use the Hough transform to to detect the angular offset of the image (ie: how much it is rotated), but you still need to detect the boundaries of the form. It also had to accommodate for the boundaries of the piece of paper itself.
This was a lucky break for me, because it basically showed a black and white image in the middle of a big black border.
Apply an aggressive, 5x5 median filter to remove some noise.
Convert from grayscale to black and white (rescale intensity values from [0,255] to [0,1]).
Calculate the Principal Component Analysis (ie: calculate the Eigenvectors of the covariance matrix for your image from the calculated Eigenvalues) (http://en.wikipedia.org/wiki/Principal_component_analysis#Derivation_of_PCA_using_the_covariance_method)
4) This gives you a basis vector. You simply use that to re-orient your image to a standard basis matrix (ie: [1,0],[0,1]).
Your image is now aligned beautifully. I did this for normalizing the orientation of MRI scans of entire human brains.
You also know that you have a massive black border around the actual image. You simply keep deleting rows from the top and bottom, and both sides of the image until they are all gone. You can temporarily apply a 7x7 median or mode filter to a copy of the image so far at this point. It helps rule out too much border remaining in the final image from thumbprints, dirt, etc.