Method for fixed pattern noise (FPN) correction
View Patent ↗A method for correcting column Fixed Pattern Noise (FPN) in an image sensor offers a compromise between speed and precision for calculating column FPN offsets. The present correction technique is digital, and is applied after the pixel signal voltages have been digitized by an ADC. A first Optical Black (OB) pixel is sampled and compared to a target level. An offset is stored, and an appropriate push-size is determined. Additional OB pixels are sampled and the offset is applied. The push-size is increased or decreased depending upon whether the pixel signal with the applied offset is above or below a target value. This new offset value is written to memory, and the push-size is reduced, and the process is repeated until the last OB pixel has been processed. The resulting offset is applied to the signal pixels in a column. The primary advantage of this approach is that within a single frame, a good estimate of the column offset is made which is not as affected by outlier pixels (such as “hot” or “flashing” pixels).
1. A method of column Fixed Pattern Noise (FPN) correction, the method comprising:
sampling a first optical black (OB) pixel level;
calculating a difference between the first OB pixel level and a target level;
storing the difference in a memory as an offset;
setting a push size;
sampling a next OB pixel level;
reading the offset from memory;
applying the offset to the next OB pixel level to form a modified next OB pixel level;
determining whether the modified next OB pixel level is above or below the target level;
if the modified next OB pixel level is above the target level, then the offset value is pushed down by a push-size and written to memory;
if the modified next OB pixel is below the target level, then the offset value is pushed up by a push-size and written to memory;
comparing the push-size with a minimum step size;
if the push-size is larger than the minimum step size, then reducing the push-size; and
determining whether a last OB pixel has been sampled, and if not, then repeating sampling a next OB pixel.
2. The method of claim 1 , wherein the target level is calculated by averaging OB pixels at a beginning of a line, which have not had any column fixed pattern noise correction applied.
3. The method of claim 1 , wherein the step of setting a push-size comprises setting the push-size to a significance parameter, wherein the significance parameter is set in proportion to the average uncertainty of a first direct offset estimate.
4. The method of claim 1 , wherein if the push-size is larger than the minimum step size, then the push-size is divided by two.
5. The method of claim 1 , wherein after a last OB pixel has been sampled, an offset value in memory is applied to each signal pixel in a column.
6. The method of claim 3 , wherein the significance parameter is scaled based on a gain setting.
7. The method of claim 1 , further comprising:
after determining that a last OB pixel has not been sampled, repeating the steps of:
sampling a next OB pixel level;
reading the offset from memory;
applying the offset to the next OB pixel level to form a modified next OB pixel level;
determining whether the modified next OB pixel level is above or below the target level;
if the modified next OB pixel level is above the target level, then the offset value is pushed down by a push-size and written to memory;
if the modified next OB pixel is below the target level, then the offset value is pushed up by a push-size and written to memory;
comparing the push-size with a minimum step size;
if the push-size is larger than the minimum step size, then reducing the push-size; and
determining whether a last OB pixel has been sampled.