Core modules of this library.
Given an image of persons face, we may want to detect the region of that persons eyes (I've hidden the definition of some plotting functions such as show and plot_images for readability).
img = cv.cvtColor(cv.imread('../images/face.jpeg'), cv.COLOR_BGR2RGB)
img = cv.resize(img, (192, 128))
show(img)
For such an image, it is useful to have a template filter that matches the size of the eye. Templates sizes can be determined by manual tuning or automated search algorithms. Below is an example of an appropriate template.
esf = EllipticalSeparabilityFilter(axes_in=(4,4), axes_out=(20,20),angle=0)
plot_images(esf.ellipse_templates())
Using this template we can caluclate a separability map from the image.
sepmap = esf(img)
show(sepmap)
Using the find_circles function, we can find a number of circles in the image.
points = esf.find_circles(img, num_circles=2)
print(points)
plt.imshow(sepmap.squeeze())
plt.scatter(*points.T, color='red', s=100)
img = cv.cvtColor(cv.imread('../images/eye.png'), cv.COLOR_BGR2RGB)
img = cv.resize(img, (192, 128))
esf = EllipticalSeparabilityFilter(axes_in=(4,4), axes_out=(20,20),angle=0)
sepmap = esf(img)
plot_images([img, *esf.ellipse_templates(), sepmap])
esf = EllipticalSeparabilityFilter(axes_in=(20,20), axes_out=(30,20),angle=0)
sepmap = esf(img)
plot_images([img, *esf.ellipse_templates(), sepmap])
esf = EllipticalSeparabilityFilter(axes_in=(10,10), axes_out=(40,20),angle=-5)
sepmap = esf(img)
plot_images([img, *esf.ellipse_templates(), sepmap])
img = cv.cvtColor(cv.imread('../images/football.jpg'), cv.COLOR_BGR2RGB)
img = cv.resize(img, (192, 128))
esf = EllipticalSeparabilityFilter(axes_in=(8,8), axes_out=(25,25),angle=0)
sepmap = esf(img)
plot_images([img, *esf.ellipse_templates(), sepmap])
esf = EllipticalSeparabilityFilter(axes_in=(3,3), axes_out=(40,40),angle=0)
sepmap = esf(img)
plot_images([img, *esf.ellipse_templates(), sepmap])
