6 look- up table (lut), 7 fixed pattern noise (fpn) correction, 8 off-line calibration – Rainbow Electronics ATMOS™ 2M60 User Manual

Page 13: 9 online correction

Advertising
background image

13

5440A–IMAGE–10/05

[Preliminary] ATMOS -2M60/2M30

5.6

Look- up Table (LUT)

The look-up Table is a conversion table which applies to the sensor data. At each input pixel
value corresponds a single output value. This allows you to apply a correction table like a
gamma correction for example. The Look-up Table does not apply to the test pattern. The Look-
up Table might be generated by the camera by programming the gamma setting:

• Gamma 0.125 to 7.875: code 1 to 63 (0 not admitted)

You might also write your own correction table via the serial communication channel. Four banks
are available to store four distinct correction tables.

See register LUT Coefficient Save in Dataflash @ 10DH, Internal Register Mapping on

page 16

.

See register LUT Coefficient Restore in Dataflash @ 10EH, Internal Register Mapping on

page 16

. See register Processing Control bit 2 @ 202H, Internal register mapping on

page 16

.

5.7

Fixed Pattern Noise (FPN) Correction

The fixed pattern noise (FPN) is a spatial fluctuation of the sensor data particularly between suc-
cessive columns. The purpose is to calculate an offset for each column of the frame in the
darkness. The table of column offsets is stored in the internal memory. If the FPN correction is
then enabled the offset values are subtracted to the value of the input pixel. Two methods are
available to realize this processing: the off-line calibration and the online correction.

5.8

Off-line Calibration

A calibration step is required while the camera is in darkness condition. Offset coefficients for
each column are calculated on an average value measured on the 1024 first lines. Offset coeffi-
cients are 8-bit data. Therefore to be efficient the pixel value should be always within the values
1LSB to 255 LSB. If not, the overflow/underflow flags are set in the status register. The calibra-
tion step must be done in full resolution. Calibration data can be saved in nonvolatile memory
and four distinct banks are reserved for this use. You might access to these banks in read or
write mode.

See register Processing Control bits [4:3] @ 202H, Internal register mapping on

page 16

.

See register Calibration Control @ 200H, Internal register mapping on

page 16

.

See register FPN Coefficient Save in Dataflash @ 10BH, Internal register mapping on

page 16

.

See register FPN Coefficient Restore in Dataflash @ 10CH, Internal Register Mapping on

page 16

.

5.9

Online Correction

No calibration step is required. When online correction is enabled the camera calculates the off-
set coefficients for each column on masked lines and then processes the input sensor data.
These coefficients are updated at each frame. The calculated values are less accurate with
these methods but unlike the off-line calibration do not depend on the settings like the analog
gain.

See register Processing Control bits [4:3] @ 202H, Internal Register Mapping on

page 16

.

See register FPN Coefficient Save in Dataflash @ 10BH,Internal Register Mapping on

page 16

.

See register FPN Coefficient Restore in Dataflash @ 10CH, Internal Register Mapping on

page 16

.

Advertising