Help please with training the classifier

Hello, please help me to improve training my classifier. What I want is just function that after step metod of vision.CascadeObjectDetector on test image, will crop from that image (that I just tested to detect any objects) all detected bounding boxes and write each detected box as a separate image to separate file and save eache of them in one directory on disk (for example in c:\test).

If I can do it then I believe I can use many of these new images as negative images to train my classifier to be more accurate. Could someone help me please with the function I described?

P.S. Below just the information from Help about this step method: BBOX = step(DETECTOR,I) returns BBOX, an M-by-4 matrix defining M bounding boxes containing the detected objects. This method performs multi-scale object detection on the input image, I.

Each row of the output matrix, BBOX, contains a four-element vector, [x y width height], that specifies in pixels, the upper left corner and size of a bounding box.

[Andrew Tim]

Can anybody help me?...

Answer 1

[Image Analyst]

See how I crop out blobs according to their bounding box in my Image Segmentation Tutorial: http://www.mathworks.com/matlabcentral/fileexchange/?term=authorid%3A31862

Answer 2

[Dima Lisin]

Try the imcrop function from the Image Processing Toolbox.

[Andrew Tim]

Thank you sir, but could you help me to write the correct code for this?

For example after setting the detector system object with: detector = vision.CascadeObjectDetector('ExampleClassifier.xml');

What code should I write to crop from image named IMG1 all detected (by detector initialized above) bounding boxes and write each detected box as a separate image to separate file and save eache of them in one directory on disk (for example in c:\test)?