Two interactions dominate what happens when X-rays pass through a sample. One of them creates the image. The other degrades it.
The photoelectric effect, briefly
In a photoelectric interaction, an X-ray photon strikes an electron and deposits all of its energy, ejecting that electron from its shell. The photon is gone. It either passed through or it did not.
That binary quality is what makes an image. The probability of photoelectric absorption depends very strongly on the atomic number of the material, which is why dense, high-atomic-number materials like solder and copper absorb heavily while low-density materials barely register. The variation in absorption across a sample is the contrast in the radiograph.
What Compton scattering does instead
Compton scattering is a partial interaction. An X-ray photon collides with an electron, typically a loosely bound outer-shell electron, and transfers only part of its energy.
The electron is knocked free, and the photon continues on with reduced energy in a new direction, deflected by some angle theta from its original path.
That change of direction is the entire problem.
The angle determines the energy transfer
How much energy the photon gives up depends on how sharply it is deflected.
A small scattering angle means very little energy transferred, and the photon carries on with nearly all of its original energy. A large angle means substantially more energy given to the electron, and the photon continues with correspondingly less.
At the extremes, a photon deflected by essentially nothing loses essentially nothing, and a photon scattered straight back the way it came gives up the maximum possible.
Why this makes images noisy
Radiographic imaging assumes photons travel in straight lines. A pixel on the detector represents what happened along the path from the source through the sample to that specific point, and that assumption is what allows the image to mean anything.
A Compton-scattered photon breaks it. It leaves the material at an angle unrelated to the straight-line path, lands somewhere on the detector it has no business landing, and contributes signal that describes nothing about the material along that line.
The visible result is exactly what you would expect from adding signal in the wrong places. Edges soften, fine detail blurs, and the image carries a general haze of noise that reduces contrast everywhere.
There is a second reason it degrades rather than merely adds. Unlike photoelectric absorption, Compton scattering is only weakly dependent on atomic number. The scattered photons therefore carry very little information about what material they passed through, so what they contribute is close to structureless. One effect encodes material differences, the other smears an approximately uniform fog across them.
Where it matters most
Both interactions are always present, but their relative importance shifts with energy.
At the low end of the range, photoelectric absorption dominates. As energy increases its contribution falls off sharply, while the Compton contribution declines much more slowly. Eventually the two are comparable, and beyond that Compton dominates.
Where that crossover happens depends on the material. Photoelectric absorption depends strongly on atomic number while Compton scattering does not, so in low-atomic-number materials like plastics and organics the crossover occurs at lower energies than in solder or copper.
The practical consequence is that inspecting thick, low-density samples, which requires higher energies to penetrate, pushes toward the regime where scatter is a larger share of what reaches the detector. Higher kV is not free, and part of what it costs is contrast.
What reduces it
Scatter cannot be eliminated, but it can be managed.
Collimation restricts the beam to the region actually being inspected, so less material is illuminated and less scatter is generated in the first place.
Geometry helps as well. Scattered photons leave the sample at angles that diverge from the direct beam, so increasing the distance between the sample and the detector gives them more opportunity to miss it entirely while direct photons continue on their intended paths. Magnification geometry, where the sample sits close to the source and well away from the detector, produces this effect as a side benefit of how it is already arranged.
The short version
Photoelectric absorption is the effect doing the useful work, because it depends on material and removes photons cleanly. Compton scattering redirects photons into the wrong places carrying little material information, which is why it appears as noise.
Every X-ray image is both happening at once, and image quality is largely a question of how favorable that ratio is for the sample being inspected.
