1) The biggest need and potential for IP and CV education is not for those preparing careers in the CVPR type professions, but rather for those in ALL professions, and even trades. We need to look seriously at how to best present needed knowledge to a more general group than those who would read the current textbooks.
We all live in 2D and 3D spaces and have a need to understand them. A large amount of information about the world is obtained from images; this applies to ALL SCIENCES as well as other areas. Images are of prime importance in communication of information.
We need to integrate education in the processing of computer images with all other study of computers. Also, computer literacy -- whatever it is -- requires some knowledge of processing digital images.
2) Much important 2D image processing is intuitive and inherently interesting, partly because of the broad native capability of the human visual system. This, together with strong applications appeal, makes IP and CV ideal topics to use in studying more general computer concepts, such as coding, data structures, and algorithms. The intuitive and iconic structure of IP problems becomes an aid to understanding other algorithms; for example, study depth-first search and recursion by via a depth-first coloring algorithm.
3) Much of 3D vision/graphics, although also having great intuitive appeal, have special technical requirements which place these topics at an advanced level with several prerequisites. Our traditional approach has been to develop materials only for students in this type of course.
SOME RECOMMENDATIONS
A) IP and CV topics should be integrated into both introductory computer
courses and higher level computing courses. Short units of materials
are needed to support one or several weeks of study/projects on the
topics below. Data structures books should include a chapter on the
2D digital image as a data structure/class of importance. (None of the
many data structures texts on my shelf have any such material; nor
is the word "image" in any of the indexes at the back!) This material
should be within the grasp of high school students.
Topics would include coding/representation, thresholding, binary images, connected components, object or defect counting, edges and regions; properties of images from moving sensors or of moving objects; and most important, an survey of applications. Coverage is dependent on goals, but enough time should be spent to appreciate (a) that many things can be done, yet (b) that almost none of our processing is perfect. Grabbing some real images in a lab and processing them would be a big plus.
B) To properly do 3D CV and graphics, mathematical and programming knowledge must be in place. There are advantages in studying both computer graphics and CV in the same course; at least for those several topics which share the same models and mathematics. We may have to combine these two areas in a BS course so that all of our students will be able to study the basics of both of these important areas.
Topics would include cameras/sensors and imaging geometry; transformations, computation of and analysis of surface features; and object modeling.
C) Lab work is critical for our field. We are at an opportune time to provide this via the Internet. Important image databases are available and so are programs. Real problems are abundant and are usually easy to understand by even beginning students (although their computer solution may not be). With the proliferation of multimedia computer equipment, many institutions will have inexpensive equipment for acquiring digital images so that students can define and develop their own projects.