Shadows exist naturally in any illuminated image, and can convey a lot of information. I combine various published shadow detection techniques to improve and optimize shadow detection in images, then associate these shadows with their respective objects in images. Following a projection of the shadow to its object will create a ray along which the light source will lie. In images with multiple objects, the intersection of the rays will be identified to pinpoint the location of the light source.
OpenCV 4.0
Skimage
Matplotlib
Numpy
CVLib
Download LABLog.py, the 'images' folder for test images, and gaborFilter.py from Object_detection into the same directory. This is the current most up-to-date set of files.
Run LABLog.py first with the command:
python3.7 LABLog.py --image ../images/[image_name]
This will output the orginial image in black and white with shadow pixels marked in black. It will also output the original image segemented with k-means.
Run detectCVLib.py with the command:
python3.7 gaborFilter.py --image ../images/[image_name]
This will output an image specified in the code with the texture and object boundaries labeled in white.
Run harrisCorner.py with the command:
python3.7 harrisCorner.py --image ../../images/[image_name]
This will output an image with all 'critical' points identified in red.
Run objectFeature.py with the command:
python3.7 objectFeature.py --image ../images/[image_name]
This will output an image with the shadow and outline of the object, along with critical points identified in each region. Points found on the object are in red, and points in the shadow region are labeled in blue.
