Repository navigation
Conversation
Uniform veiling flare on the calibration scene cannot be represented by a 3x3 matrix, so the fit previously distorted the matrix to accommodate it; measured at 1-2% of full scale on real lightbox captures, costing up to 0.7 mean delta E on the deployed matrix. The offset is fitted alongside the six matrix coefficients and then discarded - only the matrix is written. Configurable via ccm.flare_compensation (default on) in the config file, server API and Run page; the fitted flare is logged, reported in the metrics (flare_pct, per-patch de_flare) and warned about above 1.5%.
Adds a 'Flare-corrected dE' stat to the colour-accuracy card, scoring the matrix with the fitted stray-light offset discounted - the raw mean dE is scored against the flare-contaminated capture and so reads slightly worse for flare-aware fits. The de_flare/flare_pct metrics are now emitted only when flare compensation actually ran, so the stat hides for runs without it.
Add a Sharpen tab that sweeps rpi.sharpen threshold candidates over temporary tunings and measures grey-patch noise amplification on the Macbeth chart against a sharpening-disabled baseline. The smallest threshold within tolerance is recommended and can be applied to the project's generated tuning. Results persist in <project>/sharpen/, which calibration runs never scan.
Add a strength and limit sweep measuring slanted edges in the processed output: acutance gain (the CSF-weighted MTF area, which follows sharpening where optics-pinned MTF50 cannot), MTF peak, and edge halos as growth over the baseline's own profile, with edge geometry locked across sweep points. Strength is recommended as the greatest acutance gain within perceptually calibrated halo (5%) and MTF-peak (1.5) caps; the limit is then swept at that strength on high-contrast edges, recommending the largest within the cap and abstaining when the limit never engages. Also adds a perceptual noise-tolerance default and a warning when the threshold gates away the detail the sharpener needs.
…ed reference limit Scored on edge acutance alone, the strength recommendation can collapse to a weak value that draws crisp edges but soft newsprint and fabric, so the sweep now auto-selects faint-texture tiles from the baseline, reports a texture gain per point, and requires candidates to stay within 5% of the best in-cap point's texture gain. Strength points are also written with an explicit reference limit of 1.0 rather than inheriting the base tuning's, so the file's starting state can no longer steer the recommendation towards its own corner.
The tab's card intros and field hints had grown into full method descriptions; trim them to what is needed at the point of use and point at the docs. The ctt-server documentation gains a settings table per sweep explaining each parameter's default and rationale, plus a note on the fixed reference limit and texture floor.
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
1 participant
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
No description provided.