Algorithmic pixel manipulation in post — saturation, contrast, noise reduction, stabilization. Core of digital finishing. Happens in every color suite.
You're working in the edit and notice: the footage from the first day of shooting isn't consistent with the second. Colors are drifting, the lighting situation has shifted, and somewhere the camera assistant forgot to note the ISO setting. This is where image processing comes in — not as a creative last resort, but as a systematic tool in digital post-production. It operates at the pixel level, manipulating brightness values, color channels, and frequencies to bring raw footage into a consistent, visible form.
In practice, this means: you open your compositing tool — be it DaVinci Fusion, Nuke, or After Effects — and apply algorithmic filters. Adjust saturation for color consistency across scenes. Regulate contrast to preserve detail in highlights and shadows. Noise reduction for high-sensitivity material shot at ISO 6400 or higher. Stabilization for handheld footage that lacked a gimbal — the software analyzes frame-to-frame motion and removes it. The crucial point: these operations work non-destructively on the digital data. You don't lose information, but rather reorganize what the sensor has already captured.
The technical boundary lies between correction and creation. Image processing is classically correction — you fix what went wrong or was unavoidable on set. But the same techniques also enable stylistic interventions: local contrast enhancement (clarity), selective sharpening, color grading with curves and LUTs. In larger productions, the colorist works primarily with image processing; the VFX supervisor uses it for preparation for composites (keying requires clean source data).
A practical warning: noise reduction is a trade-off. Aggressive filters smooth not only noise but also fine textures and detail. A subtle hand is better here than an automatic solution. The same applies to stabilization — too aggressive, and moving objects in the background distort, the image composition becomes unstable. The best results are achieved when you use image processing as a means of support — not rescue.