User guide

Page 112 Document No: 11294 LBA-USB User Guide
4.26 Gamma Correction
If your camera has a gamma value less than or greater than 1, LBA-USB can be set to
correct for your camera’s non-linear response. Enter the gamma of the camera in the
Gamma field in the Camera... dialog box. Each pixel of each new frame of data will be
automatically corrected as defined in the equation shown below. An entry of “1.0”
disables gamma correction.
P
P
Z
z
g
×
⎟
⎠
⎞
⎜
⎝
⎛
=
/1
Where:
z = Gamma corrected pixel intensity
Z = Uncorrected pixel intensity value
g = Gamma
P = The maximum value for a pixel (255 for 8-bit
cameras, 1023 for 10-bit cameras, and 4095 for 12-
bit cameras)
Note: Be sure of the Gamma correction value. If necessary, run a response curve on
the camera. Standard published gamma values are usually averages for
particular camera types and may not always be adequate for obtaining the
desired accuracy. Also, be wary of gamma values less than 1 published for CCD
cameras. These values are usually approximations obtained by using two-piece
linear fits to an exponential gamma curve. Whenever possible use CCD cameras
which allow for a gamma setting of “1.0”.
4.27 Convolution
Convolution algorithms in LBA-USB may take on a number of forms, some of which
might not fit the exact description that is to follow. In the broadest sense, convolution
refers to a general-purpose algorithm that can be used in performing a variety of area
process transformations. One such general-purpose algorithm will be described here.
For the purpose of this description, the best way to understand a convolution is to think
of it is a weighted summation process. Each pixel in an image becomes the center
element in a neighborhood of pixels. A similarly dimensioned convolution kernel
multiplies each pixel in the neighborhood. The sum of these products is then used to
replace the center pixel.
Each element of the convolution kernel is a weighting factor called a convolution
coefficient. The size and arrangement of the convolution coefficients in a convolution
kernel determine the type of area transform that will be applied to the image data.