
Enblend 4.2 can automatically fall back to Nearest Feature Transform (NFT) even when graph-cut was requested. Its graph-cut generator supports seams that begin and end on image borders. If one image sits completely inside another, the required seam can form a closed loop, outside that documented domain. This is an algorithm limitation, not proof that the requested setting was ignored accidentally.
Key Takeaways
- NFT initially follows overlap geometry; graph-cut uses image content.
- NFT's later optimizers can still use content, so its final seam is not purely geometric.
- A contained image can require the documented automatic NFT fallback.
Separate the overlap shape from the subject
Consider a hypothetical larger photograph with a smaller registered insert entirely inside it. A seam surrounding the insert does not meet the outer image borders in the same way as a conventional side-by-side overlap.
Inspect that footprint before repeatedly changing quality settings. A closed boundary remains closed even if the subject is a simple wall. Conversely, ordinary overlapping frames may satisfy the graph-cut boundary requirement while still containing a difficult fence or person.
Understand the two generators
Chapter 5 of the official Enblend 4.2 package manual makes this distinction:
| Generator | Primary seam decision | Important boundary |
|---|---|---|
nearest-feature-transform | Shape of the overlap | Default subsequent optimizers refine using image content |
graph-cut | Content-derived cost, including dissimilarity | Needs border-to-border seams; contained-image cases can fall back |
The manual recommends graph-cut with --fine-mask; in that combination, later optimizers are automatically disabled. Graph-cut also requires more memory and computation than NFT. Those are documented tradeoffs, not measured timings for your files.
Request a diagnostic output
For an actual overlapping pair, an illustrative diagnostic invocation is:
enblend --primary-seam-generator=graph-cut --fine-mask --visualize --output=joined.tif registered-a.tif registered-b.tif
Replace the names with registered images. The visualization described in chapter 6 distinguishes the initial and final seam; it also displays overlap differences and optimization information. Read its legend rather than treating every colored line as a defect in the exported photograph. No visualization was generated for this guide.
Compare the visible route with the source footprints. For an enclosing seam, interpret NFT fallback as expected behavior for that topology instead of assuming a graph-cut failure must be repaired through stronger settings.
Fix the problem at the appropriate stage
If two recorded objects do not coincide, revisit registration or capture assumptions. A different seam generator can choose a less objectionable transition but cannot guarantee agreement between incompatible viewpoints. The panorama seam repair guide explains why inventing or duplicating structures can conceal the wrong issue.
Likewise, Boundary Warp versus crop concerns finishing the outer canvas, not choosing the internal seam algorithm. Accept a proposed blend only after inspecting both the seam region and the surviving subject geometry.
Related tasks
- How Do I Repair a Panorama Stitch Without Duplicating a Building?
- Should I Use Boundary Warp or Crop the Empty Edges of a Panorama?
Sources and scope
- Enblend 4.2 official package manual: Enblend 4.2 chapters 5-6: NFT overlap geometry with subsequent content optimizers; graph-cut content cost, fine-mask disabling optimizers, added resource cost and contained-image closed-seam automatic NFT fallback; visualization initial/final seams.
Research checked on 2026-10-23 (Asia/Shanghai). The official SourceForge archive returned HTTP 200 and its bundled 4.2 HTML manual was extracted and read locally; this is not a claim that the public manual HTML endpoint was readable. Enblend 4.2 chapters 5-6: NFT overlap geometry with subsequent content optimizers; graph-cut content cost, fine-mask disabling optimizers, added resource cost and contained-image closed-seam automatic NFT fallback; visualization initial/final seams. Commands and comparison procedures are illustrative and were not executed; the cover is decorative.
Prepared by Imagild Editorial, the company editorial team responsible for these guides. AI assisted research and drafting. This is documentation-based guidance with proposed examples, not a claim of hands-on testing or an independent product review. Report a correction.
Cover: 85 LENS (Ngô Hoàng Auth) / Pexels, Pexels License. This is an illustrative source photograph, not an application screenshot or a generated test result.