Framewise

Framewise is an adjective and adverb used to describe something that pertains to individual frames within a sequence of images or video. It is commonly used in fields such as video processing, computer vision, and neuroimaging to distinguish per-frame operations from methods that depend on longer temporal context.

In video processing and computer vision, framewise methods perform tasks independently for each frame, such as framewise normalization, per-frame feature extraction, or per-frame classification and detection. Some approaches incorporate temporal consistency through smoothing or post-processing, but the basic framewise approach treats frames as separate units, enabling simple, scalable analysis across long sequences.

In neuroimaging, framewise displacement (FD) is a standard metric for quantifying head motion between successive MRI volumes. FD is derived from the estimated six-parameter rigid-body motion and reflects how much the head moves from one frame to the next. Researchers may apply thresholds, for example around 0.5 mm, to flag high-motion frames and perform framewise scrubbing to exclude affected volumes from analysis. FD helps gauge data quality and informs preprocessing decisions related to motion artifacts.

Framewise concepts also appear in other domains, such as per-frame labeling in audiovisual data and framewise decoding in neuroscience, where the unit of analysis is a single frame rather than a temporal window. The term emphasizes per-frame analysis as opposed to methods that rely on longer temporal aggregation.