Topics of Image Processing
Prof. Wolberg received his B.S. and M.S. degrees in Electrical Engineering from Cooper Union in 1985, and his Ph.D. degree in Computer Science from Columbia University in 1990. He was an early pioneer of image morphing, and has conducted research on warping, interpolation, registration, 3D reconstruction, structure from motion, and mosaic design and rendering. Prof. Wolberg is the recipient of an NSF Presidential Young Investigator Award, CCNY Outstanding Teaching Award, and NYC Mayor’s Award for Excellence in Science and Technology. He is the author of Digital Image Warping, the first comprehensive monograph on image warping and morphing.
Image processing and computer vision are of fundamental importance to any field in which images must be enhanced, manipulated, and analyzed. They play a key role in remote sensing, medical imaging, inspection, surveillance, autonomous vehicle guidance, and more. Generally, images comprise the input and output to an image processing operation. Computer vision, on the other hand, typically operates on input images but produces a scene description or classification as the output. In practice, image processing is invoked as low-level computer vision operations, whereby the input images are filtered prior to performing high-level computer vision reasoning. Students of this course will benefit from the direct visual realization of mathematical abstractions and concepts, and learn how to implement efficient algorithms to perform these tasks.
This course introduces fundamental concepts and algorithms for image processing and computer vision. Topics include image formation, image filtering theory, image enhancement, image reconstruction, antialiasing, warping, morphing, image registration, image mosaicing, and 3D vision. The course will emphasize computational techniques for implementing useful image processing and computer vision functions. There will be a handful of programming assignments aimed at reinforcing the material covered in class.
- Image formation, perception, representation
- Point operations: thresholding, quantization, histogram manipulation
- Neighborhood operations: blurring, sharpening, edge detection, nonlinear filters
- Fourier transforms (DFT, FFT) and frequency-domain filtering
- Image reconstruction: windowed sinc functions, cubic convolution, splines
- Geometric transformations: affine/perspective/rubbersheet warps and image morphing
- Image registration and mosaicking
- Object detection and tracking
- Camera models, stereo vision
- Understand key algorithms for point, neighborhood, and geometric operations in image processing.
- Understand key algorithms for image registration, mosaicking, and 3D reconstruction in computer vision.
- Gain proficiency in implementing key algorithms using C++ with an emphasis on efficiency and clarity.
- Establish proficiency in the use of OpenCV, the leading open source computer vision library.
- Prepare students to conduct research in image processing and computer vision.
- Homework programming assignments (40%)
- Class participation (10%)
The course will be driven by lectures, Powerpoint slides, discussion, and problem solving activities exhibited by students. All five learning goals will be assessed through the homework assignments, midterm, final exam, and class participation. Learning goals (1) - (4) will contribute equally towards the final grade, with the following weights assigned to each of the assessment methods: midterm (25%), final exam (25%), programming assignments (40%), and class participation (10%). Learning goal (5) is a byproduct of the first four learning goals since the student will now have the background and experience to use existing tools and design and implement new algorithms.