|
Image processing involves treating a two-dimensional image
as the input of a system and outputting a modified image or a set of
defining parameters related to the image. Modern image processing tends to
refer to the digital domain where the color of each pixel is specified by a
string of binary digits. But many techniques are common to analog and even
optical images.
Image processing involves many transformations and techniques, usually
derived from the field of signal processing. There are standard geometric
transformations such as enlargement, size reduction, linear translation and
rotation. It is possible to modify the colors in images such as enhancing
contrasts or even transforming the image into an entirely different color
palette according to some specific mapping system. Compositions of images
are frequently conducted to merge portions from multiple images. Another
area of interest involves interpolation. Basically, images retrieved in
some contexts are sparse with missing pixels. Standard techniques involve
simply estimating the missing pixels based on the color of the nearest
known pixels. More sophisticated techniques may involve using algorithms to
judge the missing pixels usually by factoring in the relative colors of all
surrounding pixels. Techniques to align images are also quite
straightforward. Segmentation tends to involve decomposing images into
smaller sections based on some common quality such as color or light
intensity. It is possible to extend the dynamic range of photos by
combining images that have variation in light exposure. Some of the most
sophisticated techniques include morphology and flybys. Morphs involve
images literally decomposing and then re-emerging with a different look,
like a portrait in which the subject keeps changing. Flybys re-create three
dimensional imagery by rotating two dimensional landscapes around. The Holy
Grail of image processing tends to be object recognition where software is
trained to be able to recognize and categorize the parts of an image based
on colors and outlines. Authorities are particularly interested in facial
recognition technology.
Image processing is most commonly done in Matlab which allows the input of
strings of binary digits for manipulation with pre-defined commands. More
powerful software is available for larger data files and more sophisticated
applications. Many consumer-oriented products such as Photoshop have
built-in functions that allow users to edit images through a graphical user
interface. Popular features include cropping of photos to discard unwanted
areas and red eye removal which allows the darkening of eyeballs that are
distorted by exposure during photos.
There are many application areas of image processing. Perhaps the one most
familiar to most of us is in security and surveillance applications.
Regulatory authorities have streams of video feed from cameras in public
areas. It is not practical to sort through all this data manually to
identify suspicious behavior. Police and detective agencies use intelligent
software that is able to zoom in on suspicious behavior usually triggered
by sounds, the presence of packages for protracted periods of time or
clustering of many people. Image processing allows the comparison of people
on video surveillance images to suspected rogues. There have been several
successful implementation cases where criminals have been identified within
large crowds such as sports stadiums through the use of image processing
techniques.
Another critical research area is the use of image processing for medicine.
Images obtained from medicine include photographing suspected tumors,
aberrations in blood flow and fractured areas. Techniques such as magnetic
resonance imaging and computer tomography allow the generation of raw
images. Traditionally such images had to be painstakingly scoured through
by skilled practitioners who were likely to make mistakes or miss subtle
variations in the image. Image processing techniques allow the automation
of this study to identify sources of malignancy reliably and efficiently.
They enable doctors to perform guided surgery by planning their incisions
and insertions through the maze of the human body. They allow the setting
up of complicated procedures such as blasting radiation at malignant tumors
by providing complete information on the presence of both the target as
well as innocuous materials surrounding it that need to be avoided.
A third area of significance is dealing with images obtained through remote
imagery such as satellites. We are in possession of a wealth of redundant
data on the surfaces of planets but need to use image processing to
highlight areas of interest for further study. Successful techniques have
allowed scientists to judge the presence of craters, soil and atmospheric
characteristics.
|
