Handling mixed-resolution segments in long compilations

Long-form compilations often bring together footage recorded years or production cycles apart. Broadcast highlights, interviews, archive material and newly captured inserts may arrive in standard definition, 720p, 1080p or 4K, creating mixed resolution segments within a single programme. These changes can be subtle in a timeline, yet they become obvious on large screens, projectors and mobile devices when scaling, framing or sharpness shifts from one sequence to the next.

Resolution is only one part of the problem. Frame rate, codec, pixel aspect ratio, scan type, colour characteristics and audio timing may also change at editorial joins. A ReCAP-based workflow treats each section as an individually measurable media object rather than assuming that every frame in a compilation shares the same technical profile.

Australian production teams encounter this frequently when assembling AFL, cricket, rugby league or netball programmes from material captured by different crews and broadcasters. The resulting file may be produced in Melbourne or Sydney while including inserts filmed in regional centres or smaller towns. That combination, alongside variable upload capacity across Australia, makes reliable segment-level analysis especially valuable.

Why source changes accumulate

The project website describes ReCAP as an EU-funded research initiative combining content analysis with broadcast-oriented processing. Its capabilities include metadata extraction, video quality monitoring, face and logo recognition, and duplicate-content detection. Applied to a compilation, these capabilities help reveal technical changes that basic container inspection can miss.

Archive reuse is a major source of variation. An older interview may remain in 576-line material while a newly introduced sequence arrives in high definition. Rights packages can also provide different grades of the same event: a full broadcast master, a shortened edit and a web-sized derivative. Editors then combine these with fresh interviews, still images, commercials and alternate commentary tracks.

Australia’s fragmented production market adds to the mix. Public broadcasters such as the ABC and SBS, commercial networks, subscription platforms and independent production houses can each hold separate masters of related footage. A Melbourne editor might be working with a sequence supplied by a regional station where upload conditions are limited, while the final programme is destined for a national broadcaster. Sorting out those source differences can take a fair bit of time if they are documented only in email threads.

The important distinction is between a deliberate creative transition and an accidental inconsistency. A historical insert may be intended to look softer or more compressed than the surrounding material. An incorrect scale or aspect ratio, however, may result from an ingest error. ReCAP’s analytical outputs provide evidence that lets technicians tell the two apart without flattening every sequence to the same look.

Detecting changes during ingest

The first stage is to inspect the file at both container and elementary-stream levels. Container metadata provides a quick overview, while codec parsing reveals changes that occur inside the stream. ReCAP can then combine technical measurements with scene or shot-boundary analysis to identify where one profile ends and another begins.

For a long programme, analysing every frame at full quality may be unnecessarily expensive. A practical implementation can scan the timeline at a lower analysis rate, examine transition windows in greater detail and apply targeted quality probes wherever the metadata changes. This reduces processing time while preserving enough context to locate brief errors.

Picture geometry requires particular care. The project’s aspect-ratio detection note explains how pillarboxing, letterboxing and pixel-aspect mismatches can be confused with genuine framing differences. A 4:3 standard-definition source placed inside a 16:9 timeline is not the same as an anamorphic source with incompatible display metadata.

A useful output is a segment manifest showing the technical profile of every part of the programme. It might identify the opening titles as 1920×1080 at 25 frames per second, an archive insert as 720×576 interlaced material and the closing credits as 3840×2160 progressive video. Confidence scores and exact boundary times can accompany that manifest so uncertain transitions go to a human reviewer instead of being silently accepted.

Quality checks for every segment

A single score for the whole compilation would conceal more than it reveals. Two sections may share the same nominal resolution but differ greatly in bitrate, motion complexity or visible artefacts. ReCAP’s quality measures are therefore associated with individual segments and retained alongside their technical metadata.

Checks at ingest

These checks establish whether a segment matches its supplied metadata and whether its characteristics are consistent with adjacent material. They also provide a baseline for deciding whether a quality warning reflects the source itself or a new problem introduced during transfer.

Checks during review

The results should be interpreted rather than used as an automatic rejection mechanism. Some artefacts are part of the creative history: a recovered VHS clip or a low-bitrate web upload may be deliberately retained. Other artefacts indicate that a technically valid file was assembled incorrectly. Segment-level evidence gives an editor the context needed to make that decision.

Metadata can also travel with each segment through the remaining production chain. Source identifiers, measured characteristics and confidence values can be written into a sidecar file, an exchange format or the organisation’s media asset management system. That prevents a technical issue from disappearing once it leaves the ingest application.

Linking metadata, identities and duplicates

Recognising people and logos across different source formats requires more than recording their positions in one frame. ReCAP can associate detections with the relevant segment and then compare them with identifiers elsewhere in the compilation. This is useful when a presenter appears in a low-resolution interview and again in a high-definition link from the main programme.

Matching must account for scale and image quality. A face detected in a 4:3 archival crop may have a different confidence level from the same face in a clean 1080p shot. Logos may be distorted by resizing, appear against changing backgrounds or be partly obscured. The system therefore benefits from temporal context, tolerance for minor variations and a review path for uncertain matches.

Duplicate detection serves a related purpose. It can distinguish an exact file copy from a re-encoded version, an excerpt containing overlapping footage and a sequence that merely looks similar. Perceptual hashes provide a quick comparison across resolution changes, while more detailed features can confirm whether two segments represent the same underlying material.

That distinction supports editorial and rights work. Duplicate findings can flag repeated highlights, help trace which master supplied a clip and prevent multiple versions from being published without a clear source relationship. They also reduce the risk of choosing an inferior derivative merely because its filename appeared first in an asset search.

For technical processing, the metadata provides a consistent basis for decisions. An editor might retain the native resolution of each source, crop historical material once to a common frame, or deliver a normalised master with documentary source characteristics documented. The ReCAP approach makes those policies explicit instead of embedding them in undocumented scale and transcode operations.

A workable production workflow

The first step is to preserve the delivered or captured master unchanged. A working copy can then be segmented, analysed and reviewed without placing additional generations on top of the only archival copy. This is particularly important in Australia, where production teams may move large files between metropolitan facilities and regional contributors over NBN connections, fixed wireless or satellite links with uneven upload speeds.

Review copies can be lighter, but they should retain enough resolution for scaling errors, artefacts and face or logo matches to be assessed. Where a regional crew cannot upload a full-resolution file, local proxy creation can reduce transfer time while the original remains available for final quality decisions. Keeping proxies and masters linked through timecode and segment identifiers prevents review findings from being applied to the wrong sequence.

Normalisation should happen only after the source profile is understood. Converting every segment to one resolution may remove visible joins, but it can also hide genuine errors or generate new artefacts. A controlled output policy can preserve intentional differences, crop mismatched historical material and convert only where delivery specifications require it. Broader considerations for presentation and handoff are covered in media workflow themes, which can support a structured ReCAP pilot.

The next concrete step is to choose one 90-minute compilation, ingest its mezzanine without modification, run segment-level analysis across the full timeline, and review the ten boundaries with the greatest resolution, frame-rate or aspect-ratio changes.