Any raw Synthetic Aperture Radar (SAR) image has a distinct grainy texture, a “salt and pepper” look that seems like noise. This isn't random static, however. It's a phenomenon called speckle, and it's an inherent consequence of how SAR works. Unlike a camera that captures reflected sunlight, a SAR system is coherent. It sends out its own microwave pulses and records the echoes that return.

The key word is coherent imagery. Every pulse transmitted by the SAR has a consistent phase relationship with the ones before and after it. When these waves strike the ground, they don't just hit a single, smooth point. Instead, each radar pulse illuminates a patch of ground called a resolution cell. Within that cell, there could be dozens of individual objects: rocks, leaves, blades of grass, or different facets of a building.

Inside a single resolution cell, all the individual echoes returning to the antenna combine. This is where wave interference comes into play. The total signal for that pixel is the vector sum of all the small signals from the scatterers within the cell. Because the scatterers are at slightly different distances from the antenna, their reflected waves travel slightly different path lengths. These tiny differences, often smaller than the radar's wavelength, mean the waves return to the antenna with different phases.

If two waves return with their peaks aligned (in-phase), they combine to create a much stronger signal. This is called constructive interference and results in a bright pixel. If the peak of one wave aligns with the trough of another (out-of-phase), they cancel each other out. This is destructive interference, and it results in a dark pixel.

Since the arrangement of scatterers in a natural surface is effectively random, the sum of their echoes is also random. This process is often modeled as a phasor walk. Each echo is a phasor, a vector with a magnitude (amplitude) and a direction (phase). Adding them all up is like taking a series of random steps. The final distance from the starting point is the total amplitude for that pixel. One step could be large, the next small, and in any direction. The result is a pixel value that can vary dramatically from its neighbors, even if the terrain is physically uniform.

It’s crucial to understand that speckle is not the same as thermal noise. Thermal noise is truly random, additive noise generated by the sensor's electronics. It adds to the signal. Speckle, on the other hand, is deterministic but stochastic. For a given arrangement of scatterers, the resulting speckle pattern is fixed and repeatable. If you could fly the exact same SAR acquisition again, you would get the same speckle pattern.

Because speckle arises from the combination of signals, it's considered multiplicative noise. A brighter area (stronger underlying signal) will have a larger variation in speckle intensity than a darker area. The speckle's intensity scales with the signal's intensity. We can model this relationship mathematically.

This multiplicative nature is why speckle can't simply be subtracted away. It's woven into the fabric of the signal itself. Processing techniques, known as speckle filtering or multilooking, are designed to reduce its effect by averaging pixels, but this always comes at the cost of reduced spatial resolution.

Ready to check your understanding of speckle?

Understanding the origins of speckle is the first step toward interpreting SAR imagery correctly and applying the right techniques to extract meaningful information from it.