Film-base calibration
Accurate inversion starts with an accurate base. The orange film base is not just background color — it encodes the dye couplers' built-in correction mask, and a wrong base reading is the most common cause of a color cast in the finished positive.
Sampling the base
OpenEnlarge offers two strategies for measuring the base color from the clear film edge or a selected region:
- Default (robust per-channel) — the engine computes the ~95th percentile of pixel values independently for each color channel. Working at the high end of the distribution means a few dust specks or scratches on the leader cannot drag the sample low and introduce a cast. Each channel is sampled separately, so the measured base faithfully tracks the exact orange tint of this particular roll.
- Coherent mode — instead of sampling each channel independently, the engine picks a luminance band and averages a single real RGB color from pixels that fall within it. Treating all channels together means the sampled color stays on the actual orange-mask locus even when the sampled region contains a colorful scene element. Use coherent mode when the default per-channel estimate produces a subtle cross-channel cast — for instance when shooting through a subject that overlaps the edge strip.
Per roll, not per frame
The orange mask is a property of the film emulsion and the development chemistry, not of any individual frame. All frames on a roll share the same base color. For this reason, OpenEnlarge sets the base once per roll or folder and reuses it. When you develop a roll it computes this base automatically — the confidence-filtered median of each frame's auto-detected film edge across the roll — and seeds every frame from it, so no manual sampling step is required. If the automatic base looks off, recalibrate the whole roll from the Roll tab, or override a single frame with the film-edge picker in the Basic panel. Re-sampling frame by frame is rarely necessary and can introduce subtle inconsistencies if the sampled region contains subject matter rather than clear film.
If a roll looks consistently warm or cool after inversion, the first thing to check is whether the base sample was taken from truly clear, unexposed leader. A sample taken from an overexposed frame or a colored frame edge will skew every subsequent inversion on that roll.
The white point
Alongside the base color, the engine also establishes d_max — the maximum density on the roll — from the densest sampled area of clear film (the fully transparent leader or rebate). This value anchors the white point: a pixel at d_max density is mapped to paper white in the positive.
Cropping changes what the engine sees as the densest area. If you crop tightly and exclude the original clear-leader region, the engine re-analyses the remaining pixels to find d_max. Re-analyzing after a crop ensures that the highlights in the cropped image remain correctly placed — the white point does not drift. You do not need to resample the base manually after a crop; the engine handles it automatically.
Under the hood
Density formula: density is computed as D = log₁₀(base / scan) — the base is divided out of the raw scan value before the logarithm is taken. Dividing cancels the mask and leaves the net dye density for each channel.
Default sampling: for each of the three channels the engine finds the ~95th percentile value across the sampled pixels. Using a high percentile rather than the maximum protects against bright dust specks; using a per-channel value allows the sampled color to follow the exact tint of the roll without assuming a fixed orange ratio.
Coherent sampling: pixels within a narrow luminance band (typically near the high end of the luminance histogram in the sampled region) are averaged together as a single RGB triplet. Because all three channels are averaged from the same set of pixels, the resulting color lies on the orange-mask locus and avoids the per-channel independence that can produce subtle casts when colorful content appears in the sampled region.
d_max: the engine records the densest sampled area of clear film as d_max — equivalently, the pixel with the highest raw value in the clear region. After a crop, the engine re-derives d_max from whatever region of clear film remains visible, keeping the white point stable.