JPEG Quality Estimator (Quantization Tables)

How the estimate works. The JPEG standard (ITU-T T.81, Annex K) gives example quantisation tables for luminance and chrominance. The widely used libjpeg library turns a quality factor from 1 to 100 into tables by scaling those examples: below 50 the scale is 5000 divided by the quality, above it is 200 minus twice the quality, each entry is rounded and limited to 1-255. This tool recomputes the scaled tables for every quality from 1 to 100 and picks the one that differs least from the tables in your file, using the luminance table and, if present, the chrominance table.

Reading the result. An exact match means the encoder derived its tables the libjpeg way, and that quality number is a good description. A close match means small rounding differences. No match means the encoder used its own tables (common for Photoshop, camera firmware, mozjpeg or Guetzli), and the number is only the nearest standard quality, shown as an approximation. The tables themselves are displayed as 8 × 8 grids so you can compare two files.

Other details. The frame header gives dimensions, bit depth, the number of colour components and their sampling factors, which become the chroma subsampling (4:4:4, 4:2:2, 4:2:0 and so on). The compression type shows baseline, extended sequential or progressive coding. Application segments show JFIF, Exif, XMP, ICC profile and others; the metadata summary lists the camera, software, date and orientation if an Exif block exists, and whether GPS data is present without displaying coordinates.

JPEG files (.jpg, .jpeg) up to 60 MB, baseline, extended sequential and progressive, 8-bit and 12-bit, greyscale and colour. The format is decided by the file content. Lossless JPEG files have no quantisation tables, so there is nothing to estimate for them.

1. Choose a JPEG file.
2. Read the estimated quality and the sentence that explains how reliable it is.
3. Look at subsampling, coding type and the quantisation tables.
4. Check the metadata summary and the list of segments for anything unexpected, such as data after the image end.

• Finding out how hard a JPEG was compressed before you re-save it
• Matching the quality of a new image to an existing one
• Checking whether an image was probably re-compressed at a lower setting
• Seeing whether a photo is 4:2:0 or 4:4:4, and baseline or progressive
• Understanding why two JPEGs of the same size differ in file size

The estimate is not the quality slider of the program that saved the file. Programs map their sliders to tables in their own ways, files are often saved more than once, and some encoders optimise tables per image. A number for a file with custom tables is only a nearest standard quality. It describes how coarse the tables are, not how good the picture looks, and it says nothing about earlier compressions that the file went through before the last save.

Quality cannot be proven exactly from a file that was edited and re-saved. For files where the tables match exactly, the number is reliable as "the libjpeg-style quality that produces these tables"; in every other case treat it as an estimate.