manatttta's profile - activity

Identify vegetation lanes on image I have images that consist of vegetation disponsed in lines (may or may not be straig

Hello,

I have a binary mask (CV_8UC1). I want to find all POSITIVE neighbour pixels to a given pixel.

The idea is to implement something similar to GIMP's magic wand selection, in which a user selects a pixel, and the app automatically selects all positive neighbours as a pixel

How do I do this?

Can't get quite the result I was expecting (no branches). Will investigate better, thank you!!

Hello,

I have a mask image (binary mask) and I want to skeletonize it. I have some solutions for it, but all of them return some features I'd like to remove, namely small branches and circles (see the images below). How can I do this?

I want my skeleton to be a simple continuous line with no branches nor circles.

@sturkmen Thank you very much, very good algorithm!

Hello,

I know it is possible to get the skeleton of a shape in an image from either successive morphological operations or distanceTransform. I can do this via morphological operations. But How do I do this via distanceTransform?

Right now I have:

cv::Mat input = (...);
cv::Mat output;

cv::distanceTransform(input, output, ...);

Now, how do I get the skeleton from the image transform in output?

I have been investigating the cv::fitEllipse function for extracting additional data. I read the article that is cited in the documentation (Andrew W. Fitzgibbon, R.B.Fisher. A Buyer’s Guide to Conic Fitting. Proc.5th British Machine Vision Conference, Birmingham, pp. 513-522, 1995which can be found at http://www.bmva.org/bmvc/1995/bmvc-95...).

I explicitly want to extract the a = [Axx Axy Ayy ...] matrix, but the code is so optimized that I cannot identify where these values are. I'm not even sure that the function fom OpenCV 3.0 implements that algorithm. For instance, they only estimate 5 parameters, while the cited paper refers to 6 parameters.

Does anyone has an hint on that?

Alternatively, does anyone have a hint on how I can measure how good my estimated ellipse fits the model? For now, I was relying on the fact that F(a, x) = 0, where a is my ellipse parameter vector, a = [Axx Axy Ayy ...].

Hello, I have a custom warped image, with the non existing pixels set as black.

I would like to apply something similar to BORDER_REFLECT, defined by OpenCV in its wrapers (from stitching). Is tehere a lower level function to do this?

Hello, I have been trying to compile OpenCV 64bit with VS2013 EXPRESS and CUDA 6.5

When I install CUDA, it says it can't find any compatible Visual Studio version, however the website (http://docs.nvidia.com/cuda/cuda-getting-started-guide-for-microsoft-windows/#axzz3GnALRTay) says CUDA supports Cross (x86_32 on x86_64). When I try to compile (after Cmake with WITH_CUDA enabled, I get the following error

Error   2   error MSB6006: "cmd.exe" exited with code 1.    C:\Program Files (x86)\MSBuild\Microsoft.Cpp\v4.0\V120\Microsoft.CppCommon.targets  170 5   opencv_core

I can compile the CUDA examples in this version of Visual Studio.

What does support to "Cross (x86_32 on x86_64)" mean?

Is OpenCV compatible with VS2013 Express + CUDA or not? How do I do it?

Hi, I have a set of aerial images and I run them through Bundler (SfM) which gives me the rotation and translation matrices of all cameras relative to a random coordinate system.

I want to take two of those images, img0 and img1. The first one I assume is aligned with a reference frame and therefore H0 = eye(3). I want to compute H1, the Homography from img1 to img0 so that I can stitch them into a photo-mosaic,

warpPerspective(img1, img1_warped, H1)

I think this H1 should be given by

H1 = K1*R1*(Mat::eye(3, 3, CV_32F) + ((t1 - t0)*n.t() / t0.at<float>(2)))*R0.t()*K0.inv()

This is not working. The formula might be wrong or I am not correctly interpreting the data output from bundler. What am I missing?

Hello. I am working exactly on the same and have made some progresss but I am now stuck in the Blending step. I have made a new homography class estimation both rotation and translation.The PlaneWarper effectively takes translation into account after this (check the code).

But the Blender doesn't and crashes most of the time due to bad memory allocation or bad image limits computation (all due to not taking translation into account). Did you have some progress since tour question?