Image Engineering camSPECS User Manual page 23

spectral sensitivities database is small and if constraints are required. When this method is selected,
the Smooth Components checkbox is available to enable regularization, thereby reducing noise
artifacts. The Number of Components pop-up becomes active and selects the number of principal
components extracted from the reference spectral sensitivity database that will be used for
computing the spectral sensitivity of the test camera. Use the smallest value that minimizes the
difference between the computed and the Reference Spectral Sensitivities for the set of test
cameras.
•
The EVA (Eigenvector Analysis) method applies if the variance in the spectral sensitivity database is
significant, if the test camera is significantly different from those in the spectral sensitivity database,
and if constraints are required. When this method is selected, the Smooth Components checkbox
is available to enable regularization, thereby reducing noise artifacts. The Number of Components
pop-up becomes active and selects the number of eigenvectors extracted from the reference
spectral sensitivity database that will be used for computing the spectral sensitivity of the test
camera. Use the smallest value that minimizes the difference between the computed and the
Reference Spectral Sensitivities for the set of test cameras.
3. The Dark Frame Subtraction checkbox selects a dark capture taken with the camera when the lens cap
is on or by another means to ensure no light reaches the image.
4. The Use All Channels checkbox determines whether all Image and Spectrum data is used. This option
helps determine the minimum number of measurements required on the production line to achieve a
given level of performance. If unchecked, the Significance textbox becomes active, and a number
between 0 and 1 may be entered. If the relative strength of a given LED channel exceeds Significance,
its Image and Spectrum data are used, and the report LEDstrengths.txt is written. This strength is
specified for each LED channel, and if it is 0 for all three RGB color channels, it does not need to be
included in the performance data set. However, using at least 2 LED channels per RGB color channel
and at least one white is recommended. Use the smallest value that optimizes the tradeoff between
minimizing the number of measurements and the Difference between the computed and Reference
Spectral Sensitivities for the set of test cameras. More measurements ensure greater robustness,
while fewer measurements reduce total measurement time.
5. Add Spectra Folder selects the folder containing the spectral power distribution files for the desired
individual LEDs previously saved using the "Export" feature in the separate iQ-LED software. Add
Spectra selects individual spectral files rather than an entire folder. - removes an individual selection
while C clears all selections. Each entry in the Spectrum list must have a corresponding Image entry.
Creating several kinds of white illuminants using the "3) create illuminant" function of the separate iQ-
LED software is also recommended. The spectral power distributions for the custom white illuminants
can be saved from the "spectral measurement" panel in the iQ-LED software. One helpful set of whites
to include in the list are A, D65, and E, with illuminant E most beneficial for white balancing the computed
spectral sensitivities. The naming convention for the exported and any additional white spectra is
illustrated in the iQ-LED Images window. If White is checked, the corresponding Image/Spectrum pair
will be used to white-balance the computed spectral sensitivities. Illuminant E is recommended for this
Image Engineering
Page 23 of 36