Next: Image Enhancement: Spatial Methods
Up: Lectures
Previous: Noise
Last updated on Saturday, October 7, 1995 at 6:00 PM.
Gonzalez and Woods, Ch. 4 through 4.2.
Many times, we are given an image that was acquired through some
(possibly unknown) process.
This process may have been less-than-optimal, so we may want to
enhance it.
However, before enhancing it, we need to consider what the end goal is:
- Are we enhancing the image so that it can be better seen by a human?
- Are we enhancing it prior to application of a vision algorithm?
- Are we enhancing it prior to some other stage of enhancement?
Image enhancement is always task-dependent.
There are three basic ways that people normally use to enhance an image:
- Point processing (intensity transformations)
- Spatial filtering (neighborhood operations)
- Frequency-domain filtering
The remainder of this lecture deals with point processing.
The next two lectures cover
spatial filtering
and
frequency-domain filtering.
For most images that you're given, the actual pixel value is meaningless.
You hope it relates somehow to the original light intensity (or whatever
the pixel values encode), but you don't know how much each quantization
step equates to, whether the pixel values are linear with respect to
the intensity of the input, etc. This means that you're pretty much
free to do whatever you want to do with the intensity encoding.
For this lecture, we'll deal with transformations of the form
Notice that this equation involves only the value at a single
pixel--it does not involve any neighboring pixels.
For simplicity, we'll simply write
The simplest example of an intensity transformation is
which is a simple inversion of the intensities--i.e., a
negative.
Quite often, the pixels in an image only take up part of the possible
range of values.
More, more likely, they may use the entire range, but the majority of
the pixel values may lie in a narrow range.
Similar values are more difficult to discriminate, so one can
make it easier to discern subtle contrasts by stretching the values
in the range where the majority of the pixels lie.
Sometimes its useful to use intensity transformations of the form
for the intensities at the same pixel position across multiple
images.
Next: Image Enhancement: Spatial Methods
Up: Lectures
Previous: Noise
© Bryan S. Morse, 1995