Multilook

Multilook is a processing technique used in radar imaging, particularly in synthetic aperture radar (SAR), to reduce speckle noise by averaging together multiple independent observations of the same scene, called looks. Each look is formed from a subset of the data, commonly arranged in azimuth (along-track) and/or range (across-track) directions. The resulting multilook image trades radiometric stability for finer spatial resolution, since averaging across looks smooths fine details.

In practice, multilook can be performed on complex-valued SAR data or on derived intensity images. For complex data, averaging the complex samples across looks preserves phase information before computing magnitude. For intensity imagery, the pixel values or their squared magnitudes are averaged across looks. The degree of smoothing is determined by the number of looks, and speckle variance decreases roughly in proportion to 1 divided by the number of looks, improving radiometric quality at the cost of spatial detail.

Applications and considerations. Multilook images are common in SAR product generation, where reduced speckle facilitates interpretation and automatic classification. The technique is widely used in radiometric calibration, change detection, and thematic mapping. The choice of the number of looks involves a trade-off: more looks yield cleaner images but coarser resolution. Typical look numbers range from a few to a dozen or more, depending on sensor characteristics and analysis goals.

Limitations and alternatives. While multilooking improves radiometric quality, it cannot restore high-frequency spatial information lost to averaging. It may also reduce coherence in interferometric or polarimetric applications. Alternatives and complements include despeckling filters, adaptive smoothing, and multi-resolution or nonlocal techniques that aim to suppress speckle without as much loss of detail.