SPRUJ28E November 2021 – September 2024 AM68 , AM68A , TDA4AL-Q1 , TDA4VE-Q1 , TDA4VL-Q1
The LUT based DPC suffers from the drawback that each sensor die has to be calibrated for defects. The calibration process adds cost to the sensor dies, as such the preferred DPC method relies on automatically detecting the defects based on thresholds.
The OTF DPC algorithm detects defective pixels by comparing the current pixel (d0 on the left of the figure below) against its 8 neighboring pixels with the same color (d1 ~ d8 on the left of the figure below). Let’s use dmax and dmin to denote the maximum and minimum of d1 ~ d8 respectively. If the current pixel d0 is greater than the sum of dmax and a detection threshold T (i.e., d0 > dmax + T) as shown on the right of the figure below, then the current pixel is considered a defect and its value is replaced by dmax. Similarly, if d0 < dmin – T, then the current pixel is replaced by dmin.
The detection threshold T is made adaptive by calculating the local image intensity and using a programmable look-up-table. The local intensity is approximated by the average of the second largest and the second smallest pixel values among d1 ~ d8 (dmax2 and dmin2 respectively in the figure above). With this average value, the threshold T is obtained from the look-up-table with linear interpolation. The look-up-table contains the detection thresholds (U16) at the average values of 0, 512, 1024, 2048, 4096, 8192, 16384, 32768 and the corresponding slope values (S12Q8) for linear interpolation.