Page 310 - Fundamentals of Light Microscopy and Electronic Imaging

IMAGE PROCESSING WITH FILTERS 293

(a) (b) (c)

Figure 15-5
Examples of raw and flat-field-corrected CCD images. (a) Raw image. (b) Corrected image.
(c) Flat-field frame. Notice that optical faults from uneven illumination are removed.

have been read, after which the new image is displayed on the screen (Fig. 15-6). For a
megapixel image the process can take several seconds. Linear filters are used for
smoothing, sharpening, and edge enhancement. The filters menu in the software usually
provides several kernel choices for any given operation. These filters are sensitive, so
there can be some difficulty in controlling the amount of filtering, with some filters giv-
ing overfiltration, and others giving a minimal effect. Therefore, most image-processing
programs also allow you to create your own convolution matrix, in case the menu selec-
tions are inappropriate. With practice, you will learn to select a particular filter based on
the intensity gradients and size dimensions of the detail needing adjustment in the
image.

Low-Pass Filter for Blurring

This filter removes high-spatial-frequency details such as noisy pixels and sharply
defined intensity transitions at the edges of objects in the image (Fig. 15-7). It blurs by
partially leveling the values of pixels in a small pixel neighborhood. A low-pass filter
has the effect of passing or minimally altering low-spatial-frequency components—
hence its designation as a low-pass filter—and can make cosmetic improvements to
grainy, low-S/N images, but at the expense of reduced resolution.

High-Pass Filter for Sharpening

High-pass filtering differentially emphasizes fine details in an image and is an effective
way to sharpen soft, low-contrast features in an image. The effect of sharpening and the
pattern of a strong sharpening kernel are shown in Figure 15-7. Unfortunately, this filter
also emphasizes noise and can make an image look grainy.