Image Processing Reference
In-Depth Information
Sensor
AFE
Digital signal
Sensor pixel data / linear processing
Defect cor.,
Brightness cor.
WB
Demosaicking
Color conv.
Nonlinear processing
Color/tone conv.
Noise red.
Edge enhance.
TIFF
Compression
(JPEG)
Memory card
DSP
FIGURE 8.5
An example of signal processing flow.
In the following sections, each element of the processing series shown in Figure 8.5 will
be described.
8.2.1
Defect Correction, Brightness Correction
While defect-free (white and black defects) sensors are desirable, sensor cost would
be quite expensive if only perfect zero-defect sensors could be applied for imaging
systems. Therefore, sensors having defects within the correctable range of level and
number are provided for practical use. To correct the defect, the signal of the defective
pixel is substituted by that of another, normal pixel or a signal obtained by peripheral
normal pixels. In brightness correction, the signal level is adjusted for easy subsequent
processing.
8.2.2
White Balance
Adjusting white balance (WB) prevents objects that should be expressed in white from
being expressed in tints other than white using a light source. There are two ways of
doing this. One is to take spectral distribution information of the light source directly
from sensor output distribution and adjust WB at the time of photography. The other is
to adjust WB during signal processing after image capture by color distribution of the
captured signal. Since the capabilities of DSPs are now advanced, the latter method is
preferable from the aspects of imaging system size, cost, time, and effort at the time of
photography. In the latter method, the ratio between R, G, and B is adjusted based on the
assumption that summation of pixel output of an image is achromatic (gray). But this
method fails when most of the image is highly chromatic, such as a view illuminated by
a sunset, because the above assumption does not match. Therefore, most cameras have
other options for WB.
