Convert Image Format

1 VersionTransform

Colour-layout converter for Image payloads. Re-encodes each input into the layout set by color_model — BGR, BGRA, GRAY, RGB, or RGBA. Use when downstream expects a specific channel layout different from what upstream emits.

How it fits

Typical backends

Webcam frames arrive in one channel layout; the converter re-encodes them to the layout the classifier expects before inference.

HTTP image into OCR.

Caveats

I/O contractAn input already in the target layout is returned unmodified — no allocation, no copy, no precision loss.
I/O contract`RGBA -> RGB` and `BGRA -> BGR` DROP the alpha channel; transparency is lost permanently.
I/O contract`RGB -> RGBA` and `BGR -> BGRA` ADD an opaque alpha channel of `255`, not the source frame's alpha.
AccuracyConversion to a single-channel grayscale target uses weighted luma (`0.299 R + 0.587 G + 0.114 B`), not channel averaging, via color_model.
I/O contractPixel values are not rescaled or re-typed; only the channel layout and channel count change.
Hard constraintAn unrecognised source/target pair raises an unsupported conversion error mid-tick and aborts the message.
I/O contractThe output Image's `color_space` field is updated to match color_model; downstream consumers must rely on that field rather than on assumptions about upstream layout.
State lifetimecolor_model is captured once at startup; changing it requires a restart.

Versions

8436dc37latestdefault
Automated release
6/28/2026