SPRUJ28E November 2021 – September 2024 AM68 , AM68A , TDA4AL-Q1 , TDA4VE-Q1 , TDA4VL-Q1
LUT based defect pixel correction provides the highest level of accuracy since all defects can be identified perfectly without any false positives. The LUT based DPC mechanism however suffers from the limitation that the sensor has to be calibrated (each sample has to be individually calibrated) and that can add cost to the testing and qualification process. Some automotive grade sensor vendors provide the defect pixel locations already in an on chip memory and that can additionally be utilized to program the defect LUT without adding the cost. The LUT size for the DPC is constrained to 256 entries, since it is assume for a sensor of 2 -3 Mpix resolution, 256 entries are more than sufficient. Each entry in the LUT is 29 bits with 13 bits for x coordinate, 13 bits for Y coordinate and 3 bits for the method. Figure 6-67 illustrates the organization of the LUT memory
A sample table is provided below for 7 entries. The LUT is specified in left to right and top to down fashion (Raster Scan order). As such design only has to look at the current LUT pointer to be able to determine if the current location is a defect or not.
10 | 1400 | 0 |
---|---|---|
236 | 1107 | 1 |
236 | 1200 | 1 |
488 | 10 | 1 |
488 | 100 | 2 |
800 | 138 | 3 |
900 | 1000 | 1 |
900 | 1100 | 3 |
The method filed pertains to how the defect pixel is treated once it has been detected. Several possibilities arise including copying from one of the neighbors or using an average of the neighbors. The table below summarizes different methods that can be programmed for treating defect pixels.
Method Value | Functional equivalent |
---|---|
0 | Replace with Black or white dot based on config |
1 | Dout = d4 |
2 | Dout = d5 |
3 | Dout = (d4 + d5)/2 |
4 | Dout = (d2 + d7)/2 |
5 | Dout = d2 |
6 | Dout = d7 |
7 | Dout = ((d2 + d7)/2 + (d4 + d5)/2)/2 |
The pixel arrangement and numbering convention is depicted in the image below with ‘d0’ being considered the center pixel.
When the DPC method is set to ‘0’, the output is either a black or a white pixel (depending on a separate configuration register). With this method, the LUT dpc can be used in conjunction with OTF DPC in a way that the LUT programmed defects are almost always identified by the adaptive DPC. This can be used to accurately correct for all known defects and then using adaptive DPC to correct for temperature or other operating condition based defects. (Defect pixels are strongly correlated with analog gain).