I want to implement a OCR feature.
I have collected some samples and i want to use K-Nearest to implement it.
So, i use the below code to load data and initialize KNearest
KNearest knn = new KNearest;
Mat mData, mClass;
for (int i = 0; i <= 9; ++i)
{
Mat mImage = imread( FILENAME ); // the filename format is '%d.bmp', presenting a 15x15 image
Mat mFloat;
if (mImage.empty()) break; // if the file doesn't exist
mImage.convertTo(mFloat, CV_32FC1);
mData.push_back(mFloat.reshape(1, 1));
mClass.push_back( '0' + i );
}
knn->train(mData, mClass);
Then, i call the code to find best result
for (vector<Mat>::iterator it = charset.begin(); it != charset.end(); ++it)
{
Mat mFloat;
it->convertTo(mFloat, CV_32FC1); // 'it' presents a 15x15 gray image
float result = knn->find_nearest(mFloat.reshape(1, 1), knn->get_max_k());
}
But, my application crashes at find_nearest.
Anyone could help me?
I seemed to find the problem...
My sample image is a converted gray image by cvtColor, but my input image isn't.
After i add
cvtColor(mImage, mImage, COLOR_BGR2GRAY);
between
if (mImage.empty()) break;
mImage.convertTo(mFloat, CV_32FC1);
find_nearest() return a value and my application is fine.
Related
I have image as follows:
I want to detect 5 dials for processing. Hough circles is detecting all other irrelevant circles. to solve this i created a plain image and generated absolute difference with this one. It gave this image:
I drew box around it and final image is:
My code is as follows:
Mat img1 = imread(image_path1, COLOR_BGR2GRAY);
Mat img2 = imread(image_path2, COLOR_BGR2GRAY);
cv::Mat diffImage;
cv::absdiff(img2, img1, diffImage);
cv::Mat foregroundMask = cv::Mat::zeros(diffImage.rows, diffImage.cols, CV_8UC3);
float threshold = 30.0f;
float dist;
for(int j=0; j<diffImage.rows; ++j)
{
for(int i=0; i<diffImage.cols; ++i)
{
cv::Vec3b pix = diffImage.at<cv::Vec3b>(j,i);
dist = (pix[0]*pix[0] + pix[1]*pix[1] + pix[2]*pix[2]);
dist = sqrt(dist);
if(dist>threshold)
{
foregroundMask.at<unsigned char>(j,i) = 255;
}
}
}
cvtColor(diffImage,diffImage,COLOR_BGR2GRAY);
Mat1b img = diffImage.clone();
// Binarize image
Mat1b bin = img > 70;
// Find non-black points
vector<Point> points;
findNonZero(bin, points);
// Get bounding rect
Rect box = boundingRect(points);
// Draw (in color)
rectangle(img1, box, Scalar(0,255,0), 3);
// Show
imshow("Result", img1);
Now the issue is i cant compare plain image with anyother iamge of different sizes. Any pointer to right direction will be very helpful.
Regards,
Saghir A. Khatr
Edit
My plain image is as follows
I want to create a standard sample plain image which can be used with any image to detect that portion...
I am trying to copy one image to another pixel by pixel (I know there are sophisticated methods available. I am trying to solve another problem and answer to this will be useful).
This is my code:
int main()
{
Mat Img;
Img = imread("../../../stereo_images/left01.jpg");
Mat copyImg = Mat::zeros(Img.size(), CV_8U);
for(int i=0; i<Img.rows; i++){
for(int j=0; j<Img.cols; j++){
copyImg.at<uchar>(j,i) = Img.at<uchar>(j,i);
}}
namedWindow("Image", CV_WINDOW_AUTOSIZE );
imshow("Image", Img);
namedWindow("copyImage", CV_WINDOW_AUTOSIZE );
imshow("copyImage", copyImg);
waitKey(0);
return 0;
}
When I run this code in visual studio I get the following error
OpenCV Error: Assertion failed (dims <= 2 && data && (unsigned)i0 < (unsigned)si
ze.p[0] && (unsigned)(i1*DataType<_Tp>::channels) < (unsigned)(size.p[1]*channel
s()) && ((((sizeof(size_t)<<28)|0x8442211) >> ((DataType<_Tp>::depth) & ((1 << 3
) - 1))*4) & 15) == elemSize1()) in cv::Mat::at, file c:\opencv\opencv-2.4.9\ope
ncv\build\include\opencv2\core\mat.hpp, line 537
I know for fact that Img's type is CV_8U. Why does this happen ?
Thanks!
// will read in a rgb image , no matter what the content is
Img = imread("../../../stereo_images/left01.jpg");
to make it read grayscale images use:
Img = imread("../../../stereo_images/left01.jpg", CV_LOAD_IMAGE_GRAYSCALE);
then, you don't need to copy per pixel (and you should even avoid that), just use:
Mat im2 = Img.clone();
if you do per-pixel loops, watch out to get the indices right. it's row-col world here, not x,y, so it should be:
copyImg.at<uchar>(i,j) = Img.at<uchar>(i,j);
in your case
I know for fact that Img's type is CV_8U.
But CV_8U is just the image depth (8-bit U-nsigned). The type also specifies the number of channels, which is usually three. One for blue, one for green and one for red in this order as default for OpenCV. The type would be CV_8UC3 (C-hannels = 3). imread will convert even a black and white image to a 3-channel image by default. imread(filename, CV_LOAD_IMAGE_GRAYSCALE) will load a 1-channel image (CV_8UC1). But if you're not sure the easiest solution is
Mat copyImg = Mat::zeros(Img.size(), Img.type());
To access the array elements you have to know the size of it. Using .at<uchar>() on a 3-channel image will only access the first channel because you have 3*8 bit per pixel. So on a 3-channel image you have to use
copyImg.at<Vec3b>(i,j) = Img.at<Vec3b>(i,j);
where Vec3b is a cv::Vec<uchar, 3>. You should also note that the first argument of at<>(,) is the index along dim 0 which are the rows and second argument cols. Or in other words in classic 2d-xy-chart order you access a pixel with .at<>(y,x) == .at<>(Point(x,y)).
your problem is with this line :
copyImg.at<uchar>(j,i) = Img.at<uchar>(j,i);
It should be :
copyImg.at<uchar>(i,j) = Img.at<uchar>(i,j);
Note that if you want to copy image you can simply do this :
Mat copyImg = Img.clone();
I'm beginner in opencv. I have not gotten main concepts of opencv in details.
So maybe my code it's too dumb;
Out of my curiosity I want to try machine learning functions like a KNN, ANN.
I have the set of images with size 28*28 pixels. I want to do train cassifier for digit recognition. So first I need to assemble train set and train classes;
Mat train_data = Mat(rows, cols, CV_32FC1);
Mat train_classes = Mat(rows, 1, CV_32SC1);
Mat img = imread(image);
Mat float_data(1, cols, CV_32FC1);
img.convertTo(float_data, CV_32FC1);
How to fill train_data with float_data ?
I thought It was smth like this:
for (int i = 0; i < train_data.rows; ++i)
{
... // image is a string which contains next image path
image = imread(image);
img.convertTo(float_data, CV_32FC1);
for( int x = 0; x < train_data.cols; x++ ){
train_data.at<float> (i, x) = float_data.at<float>( x);;
}
}
KNearest knn;
knn.train(train_data, train_classes);
but it's of course doesn't work . . .
Please, tell me how to do it right. Or at least suggest the books for dummies :)
Mat train_data; // initially empty
Mat train_labels; // empty, too.
// for each img in the train set :
Mat img = imread("image_path");
Mat float_data;
img.convertTo(float_data, CV_32FC1); // to float
train_data.push_back( float_data.reshape(1,1) ); // add 1 row (flattened image)
train_labels.push_back( label_for_image ); // add 1 item
KNearest knn;
knn.train(train_data, train_labels);
it's all the same for other ml algos !
I'm using OpenCV to extract a subimage of a scanned document and would like to use tesseract to perform OCR over this subimage.
I found out that I can use two methods for text recognition in tesseract, but so far I wasn't able to find a working solution.
A.) How can I convert a cv::Mat into a PIX*?
(PIX* is a datatype of leptonica)
Based on vasiles code below, this is essentially my current code:
cv::Mat image = cv::imread("c:/image.png");
cv::Mat subImage = image(cv::Rect(50, 200, 300, 100));
int depth;
if(subImage.depth() == CV_8U)
depth = 8;
//other cases not considered yet
PIX* pix = pixCreateHeader(subImage.size().width, subImage.size().height, depth);
pix->data = (l_uint32*) subImage.data;
tesseract::TessBaseAPI tess;
STRING text;
if(tess.ProcessPage(pix, 0, 0, &text))
{
std::cout << text.string();
}
While it doesn't crash or anything, the OCR result still is wrong. It should recognize one word of my sample image, but instead it returns some non-readable characters.
The method PIX_HEADER doesn't exist, so I used pixCreateHeader, but it doesn't take the number of channels as an argument. So how can I set the number of channels?
B.) How can I use cv::Mat for TesseractRect() ?
Tesseract offers another method for text recognition with this signature:
char * TessBaseAPI::TesseractRect (
const UINT8 * imagedata,
int bytes_per_pixel,
int bytes_per_line,
int left,
int top,
int width,
int height
)
Currently I am using the following code, but it also returns non-readable characters (although different ones than from the code above.
char* cr = tess.TesseractRect(
subImage.data,
subImage.channels(),
subImage.channels() * subImage.size().width,
0,
0,
subImage.size().width,
subImage.size().height);
tesseract::TessBaseAPI tess;
cv::Mat sub = image(cv::Rect(50, 200, 300, 100));
tess.SetImage((uchar*)sub.data, sub.size().width, sub.size().height, sub.channels(), sub.step1());
tess.Recognize(0);
const char* out = tess.GetUTF8Text();
For Anybody using the JavaCPP presets of OpenCV/Tesseract, here is what works
Mat img = imread("file.jpg");
Mat gray = new Mat();
cvtColor(img, gray, CV_BGR2GRAY);
// api is a Tesseract client which is initialised
api.SetImage(gray.data().asBuffer(),gray.size().width(),gray.size().height(),gray.channels(),gray.size1())
cv::Mat image = cv::imread(argv[1]);
cv::Mat gray;
cv::cvtColor(image, gray, CV_BGR2GRAY);
PIX *pixS = pixCreate(gray.size().width, gray.size().height, 8);
for(int i=0; i<gray.rows; i++)
for(int j=0; j<gray.cols; j++)
pixSetPixel(pixS, j,i, (l_uint32) gray.at<uchar>(i,j));
First, make a deep copy of your subImage, so that it will be stored in a coninuous memory block:
cv::Mat subImage = image(cv::Rect(50, 200, 300, 100)).clone();
Then, init a PIX headed (I don't know how) with the correct parameters.
// ???? Put your own constructor here.
PIX* pix = new PIX_HEADER(width, height, channels, depth);
OR, create it manually:
PIX pix;
pix.width = subImage.width;
...
Then set the pix data pointer to the subImage data pointer
pix.data = subImage.data;
Finally, make sure your subImage objects does not go out of scope before you finish your work with pix.
Hi
I have read the openCV reference from this site and using the following code:
VideoCapture mCap;
Mat mcolImage, mbwImage;
mCap >> mcolImage; // New frames from the camera
cvtColor( mcolImage, mcolImage, CV_BGR2RGB);
cvtColor( mcolImage, mbwImage, CV_RGB2GRAY);
QImage colImagetmp( (uchar*)mcolImage.data, mcolImage.cols, mcolImage.rows, mcolImage.step,
QImage::Format_RGB888 ); //Colour
QImage bwImagetmp ( (uchar*)mbwImage.data, mbwImage.cols, mbwImage.rows, mbwImage.step,
QImage::Format_Indexed8); //Greyscale. I hope
ui.bwDisplay->setPixmap(QPixmap::fromImage(bwImagetmp));
ui.colDisplay->setPixmap( QPixmap::fromImage(colImagetmp ));
I'm trying to convert one of the output into greyscale. Unfortunately they're still both in colour and I can't see that I've missed a step somewhere.
Thanks for the help.
You need to explicitly set a gray color table for bwImagetmp:
QVector<QRgb> colorTable;
for (int i = 0; i < 256; i++) colorTable.push_back(qRgb(i, i, i));
bwImagetmp.setColorTable(colorTable);