1. Vision and the human eye. Rods and cones.
History of camera and image processing.
Greater
sensitivity of human vision to grayscale, but greater resolution to color.
Rods are
sensitive to light intensity, cones respond to color (3 types of cones).
Cones are
mostly in fovea, central region of retina.
2. Digital image sensors. Sampling and quantization.
Shot and thermal noise.
Sampling is
used here to refer to the lateral resolution -- how many pixels or sample
points.
Quantization
refers to finite number of gray levels or colors used to represent the signal at
each pixel.
Shot noise
results from quantum effects of light sources, usually manifest at low light
levels.
Thermal noise
is a random variation in light emission due to thermal effects.
3. Simple Image metrics. Neighbors and distances;
connectivity.
4-neighbors
are the four cells located orthogonally around a central cell (assuming
rectangular grid).
8-neighbors
are the 4-neighbors plus the cells located diagonally around a central cell.
Euclidean
distance (De) is distance between two cells measured using
Pythagoras' theorem.
4-distance (D4)
(or city-block distance) is the distance between two cells measured along
4-neighbors.
8- distance
(D8) is the distance between two cells measured along 8-neighbors.
Two pixels
p and q with values from a set V are 4-connected if q
is in the set N4(p).
Two pixels p
and q with values from a set V are 8-connected if q is in
the set N8(p).
Two pixels p
and q with values from V are m-connected if
(i) q is in N4(p), or
(ii) q is a diagonal neighbor of p and the intersection of N4(p)
and N4(q) is empty.
4. Image processing software. The prototype of free software for image processing is Image,
written at the National Institute for Health (NIH), and sometimes referred to as
NIH-Image. Since up to a few years ago, only the Macintosh
computer was easy to use for doing graphics work, NIH-Image was written for
the Mac. In addition, the program has a strong bent toward image problems
arising in microscopy. This excellent program has been maintained and augmented regularly, and
is still available as a freeware program. However, with the rise of Windows,
considerable interest has developed in producing a similar program for the PC.
First a group at the University of Texas developed a program called ImageTool
with similar capabilities to NIH-Image, then a company (Scion Corporation)
which manufactures a line of image framegrabbers developed a Windows clone of NIH-Image
called ScionImage. Both of these windows-based programs are quite useful for image processing, but
there has
not been much development recently for either of them. Currently, however, the
programmer of the original NIH-Image routines has risen up and is writing
a new image processing program, Image/J, in Java. In this way, he
hopes to have a single program that will run on both Mac and Windows
systems. These three programs for Windows: Image/J, ImageTool,
and ScionImage, all have many common basic capabilities, but differ
in some of their functions, e.g., some handle a wider range of image types (.GIF, .TIF,
.JPEG, .BMP, etc.) than the others, some have FFT capability and others do not, some are
extensible by plug-ins and others are not. In general, class work will
assume Image/J.
5. Homework assignment 1.
Last modified on January 24, 2001