Does converting an image to PDF lose quality?

The honest answer is that it should not lose anything, and that in practice it usually does, for a reason that has nothing to do with the PDF format and everything to do with how converters are written.

This is worth ten minutes because the loss is invisible on the first pass and cumulative afterwards. People convert a photo to a PDF, convert that PDF back to images, send it on, someone converts it again, and by the fourth handling the thing looks like a photocopy of a fax.

The thing almost nobody tells you: a PDF speaks JPEG

A PDF stores an image inside a stream with a named filter that says how it was compressed. The most common filter for photographs is called DCTDecode, and DCTDecode is JPEG. Not “like JPEG”, not “a JPEG variant”: the bytes in the PDF are the same compressed data that would be in a .jpg file.

So the correct way to put a JPG photo into a PDF is to copy it. Read the width and height from the file’s frame header, write the object, paste the bytes in. Nothing is decoded, nothing is compressed, the photo in the document is the photo on your disk, and the PDF is the size of your photo plus about a kilobyte of page structure.

Almost nothing does this. The easy implementation is to draw the picture onto a canvas, ask the canvas for a JPEG at some quality, and write that in. It is three lines, it works, and it silently costs a generation of compression every single time. What is a PDF goes further into what else is in the file, and the same principle explains most of it.

There are three cases where copying genuinely is not possible, and a tool that tells you about them is a tool that knows what it is doing:

• Progressive JPEGs. A progressive JPEG stores the picture in passes, so it appears blurry and sharpens as it loads. It is common for images downloaded from websites and rare for photos straight from a camera. The PDF specification defines its JPEG filter for baseline data, so a progressive file has to be re-encoded.
• CMYK photos. A JPEG from a print workflow may store ink colours instead of screen colours, which needs a colour space declaration and sometimes an inversion. Copying the bytes without carrying that produces a document with the colours inverted.
• Mirrored orientations. A photo’s EXIF can ask for the picture to be turned, which a PDF can do with its own transform, or to be turned and flipped, which it cannot. The four mirrored orientations are rare and have to be re-encoded.

What a generation of JPEG loss actually is

JPEG works by cutting the picture into 8 by 8 blocks, describing each as a mixture of frequency patterns, and then storing those coefficients coarsely. The detail it throws away is the detail the eye is worst at noticing, which is why an 85 percent JPEG looks identical to the original.

The problem comes on the second pass. The picture being compressed the second time is not the original: it is the approximation, complete with the faint block structure the first pass introduced. The second pass now has to approximate that, and it spends some of its budget encoding the artefacts of the first pass rather than the picture.

One extra generation is usually invisible. Three or four start to show as softness around edges and a faint mottle in flat areas like skies and skin. Ten is a meme.

You avoid it the same way in every context: do the work once, keep the original, and prefer tools that copy rather than re-encode.

Screenshots have the opposite problem

Everything above is about photographs. For screenshots, diagrams, logos and scans of text, the mistake runs the other way: those should never become JPEGs at all, and most converters make them one anyway.

JPEG is built on the assumption that neighbouring pixels are similar and that sharp changes are rare. A photograph fits that assumption. A screenshot is the opposite: large areas of identical colour, separated by edges one pixel wide. JPEG spends its budget smoothing exactly those edges, so text gains a grey fuzz, thin lines ring, and a screenshot that was crisp on your screen looks slightly out of focus in the document.

A PDF can store that image with deflate compression instead, which is lossless and is what a PNG uses. For flat artwork it is not only perfect, it is usually smaller than the JPEG would have been, because large areas of one colour compress superbly and photographic noise does not.

A good tool decides this per picture rather than globally: try both, keep the smaller, and never spend transparency or sharpness to save bytes. That is what “auto” should mean, and rarely does.

Resolution: the loss that is not compression

There is a second way quality disappears, and it catches more people than compression does, because it happens even in lossless tools.

Printing to PDF goes through a print pipeline. The pipeline has a page size and a resolution, and your picture is scaled to fit. A 12 megapixel photo printed to an A4 PDF is resampled down to the page, and those pixels are gone: the file now contains a page-sized image, not your photograph. This is why the built-in Print to PDF routes on Windows and Android, which are otherwise handy, are the wrong tool for anything you might want back at full size.

The related question is how many pixels you actually need, and the arithmetic is simple once you see it. A printed page wants about 300 dots per inch. An A4 page is 8.27 by 11.69 inches, so a full-page image wants roughly 2,480 by 3,508 pixels, which is 8.7 megapixels. Anything less and the printer enlarges. Anything much more is wasted on paper, though not on screen. What dpi means, and when it matters has the rest of that arithmetic, including why the number stored in a file is usually a lie.

Going the other way: PDF to images

Turning a PDF back into pictures is not an extraction, it is a rendering. A PDF page is a set of drawing instructions, so something has to run them at a resolution you pick. That has two consequences.

First, the quality is your choice rather than a property of the file. Render an A4 page at 300 dpi and you get 2,480 by 3,508 pixels, which will print. Render it at 72 and you get 595 by 842, which will not. Any tool that does not ask you is deciding for you, usually low.

Second, what you get is a picture of the page, including the text as shapes. If the page had a 200 dpi scan on it and you render at 600, you have not gained detail: you have a big picture of a small scan. Rendering above the resolution of the images inside the document only adds file size.

If what you actually want is the original images that were placed in the document, rather than pictures of the pages, that is a different operation. How to get the images out of a PDF covers both and explains which one you need.

How to tell what a converter actually did

You do not have to take anybody’s word for it, and the check takes a minute.

Compare the sizes. A PDF that copied your JPEG is the size of the JPEG plus a kilobyte or two of page structure. If your 4 MB photo became a 1.2 MB PDF, it was re-compressed, and if it became a 30 MB PDF it was stored losslessly, which is wasteful for a photograph but at least lossless.

Look at the pixels. Open the PDF and zoom to 400 percent on an area of flat colour: a sky, a wall, a piece of skin. A re-compressed photo shows faint 8 by 8 blocks that the original does not. On a screenshot, zoom into the edge of some text: a JPEG shows grey speckle beside the black letters, and a lossless copy shows clean edges.

Check the page size. If the document is exactly A4 and your picture was not A4 shaped, something has scaled it. Whether that also resampled the pixels depends on the tool, and a print pipeline always does.

How to keep the quality, in practice

Going in. Use a tool that copies JPEGs rather than re-encoding them, and that stores flat artwork losslessly. Avoid print-to-PDF for anything you might need at full size. Do not let the tool resize the picture to the paper unless you want that.

Coming out. Choose the resolution deliberately: 150 dpi for reading on a screen, 300 for printing, 600 only if you are going to crop into a detail. Use PNG for pages of text and JPEG for pages of photographs.

Both ways. Keep the originals. Every conversion is a copy, and the original is the only thing that never degrades.

The short version

The PDF format is not the thing that loses quality. The converter is. A photo can go into a PDF byte for byte, a screenshot can go in losslessly, and both come out exactly as they went in, provided the tool understands what it is holding. The two things worth checking before you use one: does it copy a JPEG rather than re-encode it, and does it let you choose the resolution rather than choosing for you.

Questions

Does saving an image as a PDF reduce its quality?

Only if the tool re-encodes it. The PDF format can hold the exact bytes of your JPG, so a well-written converter copies them in and loses nothing. A print-to-PDF pipeline, which is how the built-in routes on Windows and Android work, always re-compresses and often reduces the resolution too.

What is generation loss?

JPEG throws away detail each time it compresses. Decoding a JPEG and saving it again as a JPEG therefore loses a little more, even at high quality, because the second pass throws away detail from an image that had already been approximated. Do it repeatedly and the picture visibly degrades.

Why does my screenshot look fuzzy in the PDF?

Because it was stored as JPEG. JPEG is built for photographs and smears hard edges, which is all a screenshot is made of: text, lines and flat colour. Stored losslessly, the same screenshot is perfectly sharp in the document and often smaller as well.

Is a PDF lossless?

The container is neither lossy nor lossless: it holds whatever it was given. A PDF can store an image as JPEG, which is lossy, or with deflate compression, which is lossless, and both are normal. What matters is which one your tool chose.

Does the PDF resolution matter if I am going to print it?

Yes. A photo placed on an A4 page needs about 2,480 by 3,508 pixels to be 300 dots per inch at that size. Fewer pixels means the printer has to enlarge them, which is what makes a printed page look soft even though it looked fine on screen.

Will compressing a PDF lose quality?

Usually yes, because compressing a PDF almost always means re-compressing the images inside it, reducing their resolution, or both. It is a one-way operation: the detail does not come back when you change your mind.

Does turning a PDF back into images lose quality?

It is a new rendering rather than an extraction, so what you get depends on the resolution you ask for. Render at 300 dpi and a page comes back at printing quality; render at 72 and it does not. The original images inside the file are untouched either way.