Raoul's profile - activity

I am looking for advice from people having extensive experience with computer vision. I have a collection of ultrasonographic B&W images like the one below (without the stars and dotted line):

What I would like to do is detect the contour of a blood vessel (for example, the one highlighted by the yellow star). Of course my first step would be to define the ROI and maximize the contrast. But what would then be the best algorithm to use? Segmentation with the watershed algorithm? Something else? I am little unsettled because of the image blur...

@berak Ok, thanks! I was just wondering if accessing the OpenCV C api using cython (aka "cythonize") would allow a speed gain over the current pure python api, as cython is known to be quite efficient.

I was wondering if I could gain some speed by trying to use cython and cythonize my function calls to OpenCV. Can anyone please enlighten me?

@berak Ok, thank you! However, I seem to remember that cvMat types fed to imshow() does allow OpenGL acceleration. And it seems to be accessible to python: https://code.google.com/p/pycvgl/ so I guess I'll use that...

Reading the OpenCV documentation, I do understand that a namedWindow() created with the flag WINDOW_OPENGLwill enable OpenGL support, meaning that in C++, you could feed the imshow() function with gpu data types, such as gpu::gpuMat.

However, when using an OpenGL-enabled namedWindow()with Python, will the numpy arrays fed to imshow() be displayed as OpenGL accelerated images (i.e will be passed to the gpu)? If not, is there a way to display OpenGL accelerated images in a namedWindow() using the Python interface?

I need to revert a colormap to an array of scalars. I get the image from an OpenCV video, and it looks just like the one below:

I sampled the image using GIMP, and I therefore know that the colormap is based only on hue, in an interval from 170 to 60. I would like to know if using the pixel values of the grayscale image obtained with cv2.cvtColor(frame,cv2.cv.CV_BGR2GRAY) is valid and reflects well the hue scale of the original image.

That way, I would avoid having to rebuild the colormap in code to translate it to scalars.

Please, anyone?