Image to PDF图片转 PDF

Converting images to PDF is the one job where "just use a free website" is genuinely risky. The files people convert are almost always identity documents, payslips, bank statements, medical letters and tenancy agreements — precisely the documents you would not hand to a stranger. Yet the standard converter works by uploading your file to a server you know nothing about, in a country you did not choose, under a retention policy you did not read.

This tool has no server. The PDF is constructed in your browser and handed straight to your downloads folder.

How a PDF is actually put together

A PDF is not a compressed archive or a picture format. It is a small database of numbered objects written in plain text, plus a lookup table that says where each object begins. This tool writes all of it by hand:

• A header line declaring the version, followed by four deliberately high bytes that mark the file as binary.
• A catalog object — the root of the document — pointing at the page tree.
• A page tree listing every page in order.
• For each page: a page object carrying the /MediaBox (the physical page size), an image XObject holding the picture, and a content stream — a handful of drawing operators that place the image on the page.
• A cross reference table recording the exact byte offset of every object, and a trailer pointing back at it.

The cross reference table is the part that makes hand-written PDFs fail. Every entry must be exactly twenty bytes long and must give the true byte position of its object. If a single offset is wrong the file will not open at all. That is why the bytes are assembled once and measured as they go, rather than being estimated.

Why every image is converted to JPEG first

PDF can store an image using a filter called /DCTDecode, and DCT is exactly what JPEG uses. That means a JPEG's compressed data can be copied into the PDF byte for byte with no decoding and no re-compression — the PDF reader on the other end already knows how to unpack it. A PNG, by contrast, would need to be stored as raw pixels compressed with Flate, which would mean shipping a compression library. So every image you drop in — PNG, WebP, GIF, BMP or JPEG — is drawn onto a canvas and re-encoded as JPEG before being embedded. Transparent areas are flattened onto white, because JPEG has no alpha channel and paper has no transparency either.

Points, page sizes and the DPI confusion

PDF measures everything in points, where one point is exactly 1/72 of an inch. A4 is 210 × 297 mm, which works out to 595.28 × 841.89 pt. US Letter is 612 × 792 pt.

Here is the part that trips everybody up: the page size and the image resolution are independent. Putting a 3,000-pixel-wide photo on an A4 page does not make the page bigger — it makes the image print at about 380 dots per inch. The page size controls physical dimensions; the pixel count controls sharpness. This is why the quality setting in this tool limits the longest edge of each image: past roughly 2,000 pixels on an A4 page you are storing detail no printer will reproduce and no screen will show, and paying for it in file size.

For reference, on a full A4 page: about 1,750 px on the long edge gives 150 DPI (fine for screen reading and adequate for printing), 2,340 px gives 200 DPI, and 3,500 px gives a full 300 DPI print resolution.

Contain or cover?

Contain scales the image until it fits entirely inside the margins. Nothing is lost, but if the image and the page have different shapes you get white bands on two sides. This is what you want for documents, receipts and anything where cropping could remove information.

Cover scales the image until it fills the whole area and crops whatever hangs over the edge. This gives an edge-to-edge result with no white space, and suits photo books and presentation pages — but never use it for a document you are submitting, because it will quietly cut off the sides.

Fit image as a page size sidesteps the choice entirely: each page is made the same shape as its image, so nothing is cropped and nothing is left blank. The trade-off is that the pages may all be different sizes, which some printers and portals dislike.

Margins matter more than you would think

A 10 mm margin is a good default. Set it to 0 only if you know the file will stay digital: virtually every desk printer has an unprintable border of 3–6 mm around the sheet, so a zero-margin PDF will either lose its edges or be silently shrunk to fit. If the PDF is going to be hole-punched or bound, give the margin 20 mm.

Practical notes for submitting documents

• Check the size limit before you submit. Malaysian bank, insurance and government portals commonly cap uploads at 1 MB, 2 MB or 5 MB. If your PDF is over, switch the quality setting to Smallest and rebuild — that alone usually halves the file.
• Shoot documents in even light. Lay the page flat, avoid your own shadow, and fill the frame. A well-lit phone photo at 2,000 px beats a badly-lit one at 4,000 px every time, and compresses far better because JPEG spends most of its bytes on noise.
• Put the pages in order before you build. Reordering afterwards means editing the PDF, which is a much bigger job than clicking the arrows here.
• This produces an image-only PDF. The text inside it is a picture of text, so it cannot be searched or selected. If a form specifically demands searchable or machine-readable text, you need OCR software, not an image converter.