IdentiFrame 1.1.1

First release of IdentiFrame: FrameFusion continues under a new name, with a mode made for faces and the reading of plates and texts.

Why a new name. Up to version 1.0.1 this program was published as FrameFusion. That name turned out to be already used by a commercial product that does the same kind of work, so, to avoid any confusion with it, the program continues as IdentiFrame, with a new name, logo and icon. Nothing else is lost: the code, the method and the reports carry on from FrameFusion 1.0.1, this is the same repository (renamed: links to the old FrameFusion address lead here), and the FrameFusion 1.0.0 and 1.0.1 releases stay available. To move from FrameFusion, put IdentiFrame.exe in the folder of FrameFusion.exe and delete the old executable: config.json and the tools folder (ffmpeg) are read as before.

FrameFusion 1.0.1

FrameFusion 1.0.1

• Automatic range: every time you select the subject area, In and Out move to 150 frames before and 150 after the reference frame (within the video), also on each new selection.
• Up to 301 frames per fusion (was 300).
• Timeline ruler: labels no longer overlap when zoomed in to single frames.

FrameFusion 1.0.0

First public release: fuses several frames of a video into one enhanced image of a small detail (plate, face, sign), with a report that documents every step.

What's new in 1.0.0

• Seven-step pipeline: frame extraction, subject detection, local alignment, sub-pixel registration, multi-frame fusion, deconvolution and denoising, single frame.
• Subject area drawn on the reference frame (right-drag, or Select area): the fusion starts by itself; the subject is followed through the range by normalized cross-correlation.
• ECC registration with translation, rotation, affine or perspective model (simpler models as fallback), refined to sub-pixel precision; frames chosen by alignment quality and sharpness.
• Fusion on a ×1–×4 grid: robust mean with outlier rejection, median, or shift-and-add with normalized convolution; brightness equalization and sharpness weights.
• Restoration on the luminance only: optional bilateral denoising, Richardson-Lucy deconvolution with total-variation regularization and a PSF measured on the edges of the fused image (or chosen); auto levels, local contrast (CLAHE) and grayscale.
• Linked before/after comparison (ghost of the frames used, or the reference frame, next to the result) with zoom down to the single pixel; changing only fusion or restoration parameters re-runs the last steps in a fraction of a second.
• Saved settings: named parameter sets in config.json, defaults with one button, parameters reloaded from the report.json of an earlier fusion.
• Report in text, JSON and PDF (with the images): source SHA-256, range, reference, area, parameters, steps, per-frame table, noise estimates and SHA-256 of every image written; optional aligned frames.
• Preview, zoomable timeline and ffmpeg downloader inherited from VideoSlicer; interface and reports in English, Italian, German, French, Spanish, Dutch, Polish, Greek and Albanian, switchable at any time.

Download

FrameFusion-1.0.0-windows-x64.zip is the standalone program for Windows 10/11 (64-bit). Python is
not needed: unzip it anywhere and run FrameFusion.exe.

Updating from a previous version: replace FrameFusion.exe with the new one. Your
tools folder (ffmpeg) and config.json stay as they are.

First start

• The interface starts in English; switch language (English, Italian, German,
  French, Spanish, Dutch, Polish, Greek, Albanian) in the Language box of the
  Configuration page.
• User manual with screenshots, in every language:
  docs/manual.
• ffmpeg is not included: at first start the program offers to download the
  official Windows build into a tools folder next to the executable.
• The executable is not signed, so Windows SmartScreen may show a warning the
  first time it runs.

Verify the download

SHA-256 of FrameFusion-1.0.0-windows-x64.zip:

75b1890aa96ebcf2f2c9c4070e560e3c295eec92ad188b5f5eaf1c13071d187d

Bundled software

FrameFusion.exe contains the Python interpreter and these libraries, each under
its own license; their license files are in the licenses folder of the zip:

• customtkinter 6.0.0: MIT
• pillow 13.0.0: MIT-CMU
• numpy 2.4.6: BSD-3-Clause (bundled components under 0BSD, MIT, Zlib and CC0-1.0)
• opencv-python-headless 4.14.0.94: Apache-2.0; the FFmpeg library it ships (opencv_videoio_ffmpeg*.dll) is under the LGPL-2.1: see LICENSE-3RD-PARTY.txt
• darkdetect 0.8.0: BSD-3-Clause
• packaging 26.3: Apache-2.0 OR BSD-2-Clause
• fpdf2 2.8.9: LGPL-3.0-only
• fonttools 4.67.0: MIT
• defusedxml 0.7.1: PSF-2.0
• Python 3.11.9: PSF-2.0

License

Free software under the GNU General Public License version 3 with additional
terms (author attribution): see LICENSE and NOTICE, included in the zip
together with the licenses folder (third-party license files).

Author: Andrea Cumini — www.osintinfo.net — andrea@osintinfo.net