The other day, I bumped into an old friend I had not seen in years. He looked familiar. Yet, he had also changed considerably over the years. I had to do a double take before I was sure that it was him.
Whenever we see something unexpected, and have to look again to clarify, we say that we do a “double take”.
In imaging, sometimes it is also useful to do a double take.
Many years ago, at the Image Sensors Workshop in Lake Tahoe, I presented the concept of using two rows of pixels to enhance the signal to noise ratio (SNR) of line scan imagers. We called it dual line scan. The concept was simple. Instead of using one row of pixels to scan, use two adjacent rows to scan twice. The extra signal from the redundant scan improves the Signal-to-Noise Ratio (SNR).
Since then, numerous dual line scan imagers have successfully been introduced into the market. Now, there are tri and quad line scans.
Redundancy is a powerful thing. Yet, in multiline imagers, people had only used it to improve SNR. We have not harnessed the full potential of redundant imaging. Today, I will introduce one of these untapped potentials.
Whether you sense it or not, we are constantly being exposed to low levels of background radiation. A small percentage of the materials around us are naturally radioactive. Some of the radioactive particles from outer space are not filtered by the earth’s protective fields.
Image sensors sense this radiation. In silicon, radioactive particles create a cloud of electrons that usually show up as streaks in an area image and as spikes in line scan images.
These spikes are unwelcome because they result in false positives. For example, in a line scan camera that is scanning a sheet of material for defects, it is often impossible to distinguish between a real material defect and a phantom one triggered by background radiation. Needless to say, these false positives can become quite expensive, as manufacturers either have to scrap materials or re-inspect.
In multiline imagers, we can use the redundancy to distinguish between real and false defects.
When a radioactive particle hits a dual line scan imager, it creates a spike in only one of the two lines of signal, even if the radioactive particle were to enter the imager at a point that is exactly halfway between the two rows of pixels. Here’s why. In the above scenario, the two rows of pixels will collect almost exactly the same number of electrons from the electron cloud formed. However, at the very moment that the electron cloud has formed, the two rows of pixels are scanning different regions of the object. When we reconstruct the image from the output signals of a dual line scan sensor, we have to combine signals that scan the same region of the object. The two rows of pixels scan the same region of the object at different points in time. Because the radioactive particle has corrupted the signal at only one point in time, we can still reconstruct the image using the other uncorrupted signal.
How do we know which of the redundant signals is uncorrupted? Easy. The smaller signal is always the one that is uncorrupted by background radiation.
To eliminate false defects from background radiation, the imaging system merely has to throw away the larger of the two signals that will otherwise have been averaged, if the difference between the two signals exceeds a certain threshold.
Redundancy is a powerful thing. Beyond recognizing old friends and detecting background radiation, can you think of other applications?