Convert ORF to PNG

JPEG divides the picture into 8 by 8 blocks and throws away the highest-frequency information in each one first, because that is what the eye notices least. It is a good decision for a photograph being looked at, and a poor one for a photograph being examined. The high frequencies are the hairs on a thorax, the scale rows on a wing, the serration on a leaf margin, the fine barring on a covert.

At normal size the loss is invisible. At the 300 or 400 per cent crop an identification actually happens at, it is not: fine texture smears into a soft wash, and the block boundaries themselves can appear as faint edges that were never on the animal. Neither effect is dramatic, and both are enough to turn a determinable record into a probable one. A lossless file removes the question entirely, which is worth more than the megabytes it costs.

Cameras embed more than one JPEG: a small one for the file browser, sometimes a mid-size one, and the full-resolution view the rear screen shows. This converter finds all of them and keeps the largest by pixel count rather than by file size.

The distinction is not academic. A heavily compressed full-resolution preview can occupy fewer bytes than a lightly compressed mid-size one, so choosing by file size would hand you the smaller picture while looking like the sensible rule. Anything below 640 by 480 is ignored altogether, on the grounds that a thumbnail is a worse outcome than a clear failure. Convert one frame and check the dimensions: 5,184 by 3,888 is a standard Micro Four Thirds capture, and anything less means the file has been through something other than the camera.

This is the sentence a recorder should read before relying on the file. The PNG stores what it is given, exactly and permanently. What it is given is the JPEG the camera wrote at the moment of capture, and that JPEG has already been through the compression described above, at whatever quality setting the body uses for its embedded preview.

So the correct description of the output is a lossless copy of a compressed picture. It is markedly better than exporting a fresh JPEG, which would add a second generation, and it is not the same as a file developed from the sensor data. Where a determination turns on detail at the pixel level, develop the ORF properly and export from there — and either way, note in the submission how the file was produced, because a reviewer assessing marginal detail is entitled to know.

A standard OM System or Olympus frame is 5,184 by 3,888 pixels. A bird filling a quarter of the frame is therefore about 1,300 pixels across, which is plenty for most features. A moth at life size on a 60mm macro lens can be resolved to well under a tenth of a millimetre per pixel, which is why the system is so widely used for that work.

High Res Shot changes the arithmetic where it is usable — a static subject, a tripod or a very steady hand — producing frames of around 8,160 by 6,120 or more. Whether that resolution reaches the PNG depends on what the camera embedded, so convert one and read the dimensions rather than assuming. For a live insect or a bird, the standard frame is the realistic case and it is a good one.

A 20-megapixel frame is 60 MB of raw pixel data before any compression, and PNG works by finding repetition, which a photograph contains very little of. Expect somewhere between 28 and 48 MB depending on the subject: a bird against a plain sky compresses well, a moth on lichen or a plant in grass compresses hardly at all.

That is a large attachment for an email and a large upload for a records portal, and the answer is usually not to compress but to send fewer files. Submit the full-frame PNG and one lossless crop rather than a set of twelve, and use a file transfer link where the portal allows it. Reducing the resolution to fit a limit defeats the purpose of the exercise.

PNG can store an image as an indexed palette of up to 256 colours, which is how it manages such small files for screenshots and line art. Nothing here does that: the encoder writes full 24-bit colour with eight bits per channel, so no colour in the picture is approximated by the nearest entry in a table.

That is worth stating because the alternative formats a submission form often lists include GIF, which is palette-only and reduces a photograph to 256 colours with visible banding. For a subtle green or a gradation across a wing, that reduction can be the difference between two species. PNG is the format on those lists that does not do it.

PNG supports 16 bits per channel and ORF stores 12 — two below the 14 most other raw formats write, which is a quarter of the tonal levels and a real difference on a smaller sensor. Neither number reaches the output: the embedded preview is an eight-bit JPEG, so the PNG is eight bits.

For examining a photograph this is almost never the constraint, because a reviewer is looking at detail rather than at tonal range. It becomes the constraint if the picture needs heavy brightening before the feature is visible — a moth photographed in poor light, or an underwing in shadow — since eight bits will band where a 16-bit development would not. That case wants the ORF and a raw developer, not this page.

The conversion decodes and re-encodes, so nothing from the metadata block survives: no camera, no lens, no exposure settings, no capture date, and no GPS coordinate from the body or a paired phone.

For wildlife recording that cuts both ways and both matter. The date and grid reference are the record — they belong in the submission form and in your own notes, not only inside a file, and you should read them out of the ORF before you rely on the PNG. On the other hand, a picture of a nest, a roost, an orchid or a rare fungus that no longer carries its coordinates is a photograph you can share on a forum without publishing the site, which is a habit worth having.

Recorders generally want two things: the detail that supports the determination, and enough context to believe it. Convert the full frame to PNG, then make the crop from that PNG rather than from a JPEG, and save the crop as a PNG too. Both files are then lossless copies of the same original with nothing added between them.

Resist the urge to sharpen or adjust before submitting. A reviewer assessing a marginal feature needs to know that what they are looking at came off the sensor rather than out of a slider, and an unsharpened file with a note about the conditions is worth more than a processed one that looks better. If you do adjust, say so.

Everything runs in the browser: the ORF is read from disk, scanned for its embedded pictures, decoded and written back out as a PNG. There is nothing to install, which is the practical case for a field laptop or a borrowed machine, and no account to make.

Nothing is transmitted either, which the network tab shows and a disconnected machine proves. For records of sensitive species that is the whole argument — a file that knows where a rare orchid grows should not be passing through an unnamed server on its way to being submitted, and here it does not.