Jdban's profile - activity

Thanks! So basically, its blotchy because in the mostly grayimage there isnt much variance, and when there IS variance, its a big variance in weird blotches.

I ran a couple more images through my HSV and they look less blotchy. Images with more color:

http://i.imgur.com/qDHajOT.png

http://i.imgur.com/u7Mobtj.png

Yes, I am using imread. So I changed it to use BGR2HSV instead of RGB2HSV as you suggested.

Now it looks like this (BGR to HSV): http://i.imgur.com/2oB9NHn.jpg

Instead of this (RGB TO HSV): http://i.imgur.com/Idyoy3N.jpg

It looks better, but still blocky. Is that expected?

Here is my input followed by its hsv converted output. If you look at the full size image, you can see that my hsv image is blocky and weird looking.

Here is the hue channel alone, you can see how poorly blocky it is here as well:

Is the blotchiness supposed to happen?

This is essentially all I'm doing in code:

Mat hsvImg;
cvtColor(srcImg, hsvImg, CV_RGB2HSV);
Mat channels[3];
split(hsvImg, channels);
imshow("Original", srcImg);
imshow("HSVImg", hsvImg);

When I google to see HSV images, I see stuff like this:

All the sample images I see don't look as insanely blocky as mine do. Am I doing something wrong?

Is anyone using CLion IDE? Are there any guides for CLion with Opencv?

I'm currently developing in windows using C++. Are there any CLion guides?

I'm trying to do an operation on every pixel (brightness and contrast) and am wondering if I'm missing something. Is there no way to do it smartly in the java?

I'd love my code to look like this:

byte[imagesize*pixels] imagepixels = mRgba.toPixels();

for(int k = 0; k < imagepixels.size; k++
{
   imagepixels[k] = imagepixels[k]*alpha + beta
}

mRgba.put(imagepixels);

But instead it looks like this:

    for( int y = 0;  y < mRgba.rows(); y++)
    {
        for(int x = 0; x < mRgba.cols(); x++)
        {

                dbl = mRgba.get(y, x);
                pix[0] = roundProperly(contrast*(dbl[0]) + brightness);
                pix[1] = roundProperly(contrast*(dbl[1]) + brightness);
                pix[2] = roundProperly(contrast*(dbl[2]) + brightness);
                pix[3] = 0;
                mRgba.put(y, x, pix);
                dbl = null;
        }
    }

There are tons of memory issues because dbl is being malloced extremely frequently and making the garbage collector block, and I'm doing WAY more JNI calls than I'd like.

I plan on porting this into a single C call, but I'm curious to see the straight java performance of this vs the C performance, and I'd prefer there to be a way to do the java this inefficiently.

I have an image of a crosswalk, and its very hard to do processing on it to detect the crosswalk. I was playing in paint.net and stumbled upon changing the contrast/brightness, I was able to EASILY distinguish the crosswalk so that I could threshold with HSV values. Btw I am running this on my android phone.

Here is my before image:

And here is my after image from paint.net after adjusting brightness:

In other images where its even more poorly colored, this does a great job too.

I was wondering how I could replicate this effect in OpenCV, what options are there?

Yes, that's exactly what I meant, thank you!

I'd love to see the C code and how its optimized, but it seems to be bundled inside the OpenCV Manager app and not available inside the source that I can find on SourceForge or in the Git repository. I can find the names of the function calls that call into the native code, but I can't find the actual native code anywhere. Is this code available? If so, where can I find it?

It took me FOREVER to find out how to fix this... I wish when I downloaded the SDK today it didn't have the .settings folder.

When I draw lines on android, I'm getting weird graphical glitches like so:

these lines are being drawn to the image pretty simply:

Core.line(mRgba, pt1, pt2, new Scalar(0, 255, 0), 3);

Does anyone know why its not doing what it should be doing and just putting that entire line as green?

I'm trying to find the median, which I believe can be done by using the code given here: http://tech.dir.groups.yahoo.com/group/OpenCV/message/23809

What I have so far in java is the histogram calculation working fine (I assume), and at that point, I don't know what to put insude my calcMedianOfHist function. The issue I'm running into is that I don't know how to access the bins of the histogram. The closest I've come to finding out how to do this is looking at this example for C code, but that doesn't apply to the Java. (http://docs.opencv.org/2.4.3rc/doc/tutorials/imgproc/histograms/histogram_calculation/histogram_calculation.html)

The method there says to do hist.at<float>(num), but I cannot do that in java.

I'm also not 100% sure if I should be normalizing the hist like this or not.

Any help is appreciated, thank you.

        ArrayList<Mat> list = new ArrayList<Mat>();
        list.add(mGraySubmat);

        MatOfInt one = new MatOfInt(0);
        int median = 0;

        hist = new Mat();

        MatOfInt histSize = new MatOfInt(25);
        MatOfFloat range = new MatOfFloat(0f, 256f);

        Imgproc.calcHist(Arrays.asList(mGraySubmat), one, new Mat(), hist, histSize, range);

        Core.normalize(hist, hist);

        median = calcMedianOfHist(hist, mGraySubmat.cols(), mGraySubmat.rows());

        System.out.println("Median is: " + median);