Confirm Midra fit and mask rasterization with a firmware engineer #4

Open
opened 2026-11-16 13:56:53 +01:00 by Dandaedre · 0 comments
Owner

Obtain authoritative Midra crop, fit and mask rasterization rules from an Analog Way firmware or video-processing engineer. Current continuous crop-first geometry and nearest-integer device fields are working assumptions, not a verified hardware raster model.

This is a distant-future precision follow-up, not a prerequisite for using the existing fit modes. Operator pixel measurements are not required: the available equipment cannot establish subpixel rendering, and API readback alone cannot prove physical output.

Questions and evidence

  • Confirm normalized crop units, their coordinate stage and when retained dimensions are quantized.
  • Confirm crop-first fitting for 1:1, Centered, Fullscreen and Cropped, including fractional retained sizes.
  • Establish pixel-center conventions, half-pixel tie-breaking, interpolation, negative centers and clipping/transparent-edge rules.
  • Establish mask coordinates relative to container/fitted image, composition with crop/fit, fractional boundary rounding and inclusive/exclusive edges, including fully blank axes.
  • Identify any Preview/Program, model, firmware or scaler differences.
  • Obtain rendered reference images, expected pixel bounds or an instrumented simulator rather than asking operators for unavailable measurements.

Acceptance criteria

  • Record the authoritative answers with model/firmware scope and provenance.
  • Cover odd-sized containers, asymmetric crops/masks, smaller and larger 1:1 sources, negative coordinates and both banks using the concrete cases in the engineering brief.
  • Add sanitized reference fixtures and adjust adapter quantization/rendering only where the evidence requires it.
  • Keep software math tests, protocol readback and physical-output validation explicitly distinct; document remaining uncertainties.
Obtain authoritative Midra crop, fit and mask rasterization rules from an Analog Way firmware or video-processing engineer. Current continuous crop-first geometry and nearest-integer device fields are working assumptions, not a verified hardware raster model. This is a distant-future precision follow-up, not a prerequisite for using the existing fit modes. Operator pixel measurements are not required: the available equipment cannot establish subpixel rendering, and API readback alone cannot prove physical output. ## Questions and evidence - [ ] Confirm normalized crop units, their coordinate stage and when retained dimensions are quantized. - [ ] Confirm crop-first fitting for 1:1, Centered, Fullscreen and Cropped, including fractional retained sizes. - [ ] Establish pixel-center conventions, half-pixel tie-breaking, interpolation, negative centers and clipping/transparent-edge rules. - [ ] Establish mask coordinates relative to container/fitted image, composition with crop/fit, fractional boundary rounding and inclusive/exclusive edges, including fully blank axes. - [ ] Identify any Preview/Program, model, firmware or scaler differences. - [ ] Obtain rendered reference images, expected pixel bounds or an instrumented simulator rather than asking operators for unavailable measurements. ## Acceptance criteria - [ ] Record the authoritative answers with model/firmware scope and provenance. - [ ] Cover odd-sized containers, asymmetric crops/masks, smaller and larger 1:1 sources, negative coordinates and both banks using the concrete cases in the [engineering brief](https://proxy.dandaedre.com/gitea/idle-org/canvas-pilot/src/branch/main/docs/canvas-pilot/midra-fit-precision.md). - [ ] Add sanitized reference fixtures and adjust adapter quantization/rendering only where the evidence requires it. - [ ] Keep software math tests, protocol readback and physical-output validation explicitly distinct; document remaining uncertainties.
Dandaedre added the
help wanted
label 2026-11-16 14:03:05 +01:00
Dandaedre self-assigned this 2026-11-16 14:03:11 +01:00
Dandaedre added the
wontfix
label 2026-11-18 10:11:29 +01:00
Sign in to join this conversation.
No Milestone
No project
No Assignees
1 Participants
Notifications
Due Date
The due date is invalid or out of range. Please use the format 'yyyy-mm-dd'.

No due date set.

Dependencies

No dependencies set.

Reference: idle-org/canvas-pilot#4
No description provided.