2022/04/22DSNU refers to the variation in the background signal of a camera image when no light (or photons) are incident on the sensor. It indicates how much the pixel offsets vary across the sensor when capturing an image in complete darkness.
DSNU becomes particularly important in low-light imaging, where the camera produces a dark or bias image in the absence of light. In such cases, the offset value of pixels may vary, and DSNU becomes important, especially when it is comparable to or greater than the camera's read noise (typically 1-3e-).
1. Why DSNU Matters in Imaging
1) Low-light Imaging:
In the absence of light, a camera’s sensor produces a signal that is not zero but typically an offset value (e.g., 100 grey levels). This offset is influenced by the camera’s electronic noise, but there might be variation in the offset values from one pixel to another. This variation in pixel offset values is referred to as Fixed Pattern Noise, which arises due to inconsistency in the sensor's dark signal across pixels.
2) Fixed Pattern Noise:
DSNU quantifies the time-independent variation in these offset values. It provides a measure of how consistent the sensor's dark signal is across different pixels.
2. DSNU and Image Quality
1) Standard DSNU Values:
For most low-light imaging cameras, DSNU values are usually below 0.5e- (electrons), which means that for medium or high-light conditions, the contribution of DSNU to noise is negligible.
2) Impact on Image Quality:
For images with high photon counts (e.g., hundreds or thousands of photons per pixel), DSNU has a minimal effect. For low-light applications, DSNU remains significant when it is comparable to or exceeds the camera's read noise (typically 1-3e-), which may affect image quality.
3. Limitations of DSNU:
DSNU (Dark Signal Non-Uniformity) is a tool used to quantify fixed-pattern noise, but it does not capture all types of fixed-pattern noise, such as structured noise (Fixed Pattern Noise) and time-dependent dark noise (Temporal Dark Noise).
1) Structured Patterns and Column Noise:
Structured Pattern Noise, particularly Column Noise, is a type of fixed-pattern noise that occurs when certain pixels or groups of pixels exhibit systematic differences in their dark signal (offsets), typically aligned along specific columns or rows. This noise appears in a structured pattern rather than random fluctuations.
2) Time-Dependent Variations:
Time-dependent noise refers to variations in the sensor's dark signal over time, caused by factors such as temperature fluctuations, electronic instability, or sensor aging. These variations cause the offset values of pixels to fluctuate from one exposure to another. DSNU measures time-independent noise, meaning it does not account for these changes in offset values over time. To observe time-dependent variations, a sequence of bias images (images taken with no light) captured over time is required to detect these fluctuations.
4. DSNU in Practical Applications
DSNU is negligible in high-light imaging, since the photon signal is much stronger than any dark signal variation.
For applications like single-molecule imaging, quantum imaging, and astronomical observations, cameras are required to capture extremely weak light signals. These signals are typically in the 1-3 electron (e-) level or even lower, so any additional noise (such as DSNU) can affect the quality of the final image and reduce the signal-to-noise ratio (SNR). This is why modern high-sensitivity scientific cameras have DSNU correction capabilities. The lower the DSNU value, the higher the quantitative accuracy.
Modern industrial inspection demands ever-increasing precision, especially in the semiconductor industry, where the quantitative requirements for defect signals are comparable to those in scientific low-light imaging applications. DSNU (Dark Signal Non-Uniformity) is equally important in this context. The article "Why DSNU/PRNU Correction Matters in Semiconductor Inspection" provides a very detailed explanation.
