Page 200 - The Definitive Guide to Building Java Robots

Preston_5564C06.fm Page 181 Friday, September 23, 2005 5:13 AM

CHAPTER 6 ■ VISION 181

6.2 Basic Image Processing

There are many types of image processing. I would like to classify the processing into three
types: pixel processing, area processing, and convolution and combination processing.
Pixel processing is the process of moving through an image pixel by pixel, getting a color
value, and doing something with it. Many times, you’ll need to remember values and place
them in storage arrays. For example, take Figure 6-8. There are nine pixels. If I move from left
to right and top to bottom, PixelProcessing would look at each of the color components and do
something with them. So, if I wanted to turn them to grey, the pixel at (1,0) would have an RGB
value of (255 + 0 + 0) / 3 = 85, and the resultant color would be (85,85,85).

Figure 6-8. A pixel-by-pixel image

The second type is convolution. This is the process of moving through an image pixel by
pixel, and then doing something with it in relation to its surrounding pixels. This is done via an
operator called a kernel. A kernel is a matrix of numbers that specify how to change the value
of a single pixel as a function of its surrounding pixels. You can think of a kernel as a template
that fits over each pixel and then changes it based on the values inside it. A sample matrix for
smoothing an image would be
{1/9, 1/9, 1/9,
1/9, 1/9, 1/9,
1/9, 1/9, 1/9}
The calculation process moves like this. Let’s start at the fifth (or center) pixel. Multiply the
pixel at its top left by 1/9, then move to the pixel to the right of that, which is above the center