A hands-on test with a three-page A4 document produced a counterintuitive result: despite both pages having roughly 81% white background, the PNG file on the chart page was 40% smaller than a JPG at quality 92, while the PNG on the text page was 5% larger. We found that when converting documents to images, the real decision is not which format to apply universally, but how to match each page’s content with the compression efficiency of the encoder.

What this is

This is an experiment focused on the file size of PDFs converted into images. The author rendered three pages containing text, charts and photographs at 216 DPI, then saved them in formats including PNG, JPG and WebP. The results show that JPG processes images in 8×8-pixel blocks, so dense details such as the strokes of Chinese characters can easily trigger lossy compression. PNG uses row filtering and DEFLATE compression to encode repeated content, making it more effective for large solid-color and continuous areas.

But having more white background does not necessarily make a file easier to compress. Although text pages contain fewer colors, their strokes break solid areas into smaller fragments. The distribution of color blocks on a chart page, by contrast, is better suited to PNG. Merely changing the PNG compression level for the same set of pixels produced a file-size difference of 1.8×, demonstrating that format selection and encoding parameters must be considered together.

Industry view

For document previews, contract archiving and knowledge-base development, the value of this test is that it corrects a common default: measure first, then select a format, rather than imposing a blanket rule of “PNG for text and JPG for photographs.” A more realistic workflow is to identify the proportion of solid-color areas and photographic regions, test lossless WebP, PNG and JPG separately, and record image quality, file size and conversion time.

We should also retain the objections. The test covers only one self-produced PDF, with each setting run once. It does not include scans, handwritten documents or CMYK print files, and it provides no repeated sampling. The article’s analysis of ringing is an explanation of the underlying mechanism rather than end-to-end validation. PSNR, a metric for measuring the average error between a reconstructed image and the original, also cannot replace visual judgment. The results are therefore useful for informing strategy, but not sufficient to establish a universal standard.

Impact on regular people

For enterprise IT: When processing PDFs in bulk, selecting formats by page type may reduce both storage consumption and transfer costs. Before deployment, teams should establish benchmarks using real business documents.

For individual professionals: When converting contracts, reports and presentation materials, there is no need to insist on PNG’s “lossless” label. Prioritize the final file size and the sharpness of text edges, exporting and comparing the results separately when necessary.

For the consumer market: If conversion tools adopt page-level detection or intelligent compression, users are more likely to receive reasonably sized images with clear text, but this may also increase processing time and create room for higher product pricing.