Dictionary

What Is An Encoding Process?

Encoding Process names the method a JPEG file used to compress its picture data, such as baseline or progressive. It is read from the image data itself.

encoding-process

An encoding process is the method that turns a picture into compressed image data. In a metadata report, Encoding Process names the method that a JPG file used, for example Baseline DCT, Huffman coding.

The value has 2 parts. The first part names how the picture data was compressed. The second part names how the result was packed into bits.

Which files carry it

Encoding Process is a JPEG field, and every JPEG image has it. The JPEG standard puts the information in a short header that comes before the compressed picture data.

Where the value comes from

No camera or program writes Encoding Process as a separate tag. A reader such as ExifTool works out the value from the Start of Frame marker, which opens that header. The JPEG standard defines a different marker for each encoding process.

The same header also stores the image width and height, the bits per sample, and the number of color components. ExifTool also works out the chroma subsampling from this header.

Because the value describes how the image data is stored, you cannot edit it like a normal tag. Encoding Process changes only when a program saves the picture again with other settings.

What the values look like

ExifTool shows one of 13 values. According to the Library of Congress, the common ones are baseline and progressive.

  • Baseline DCT, Huffman coding
  • Extended sequential DCT, Huffman coding
  • Progressive DCT, Huffman coding
  • Lossless, Huffman coding
  • Sequential DCT, differential Huffman coding
  • Progressive DCT, differential Huffman coding
  • Lossless, Differential Huffman coding
  • Extended sequential DCT, arithmetic coding
  • Progressive DCT, arithmetic coding
  • Lossless, arithmetic coding
  • Sequential DCT, differential arithmetic coding
  • Progressive DCT, differential arithmetic coding
  • Lossless, differential arithmetic coding

What each part of the value means

  • DCT stands for discrete cosine transform. The JPEG processes based on the DCT are lossy: they drop some detail to make the file smaller.
  • Baseline is the basic DCT process. It uses 8 bits per sample and Huffman coding.
  • Sequential means that each color component is stored in a single scan.
  • Extended sequential works like baseline, but it also allows 12 bits per sample and arithmetic coding.
  • Progressive stores the picture in several scans. Each scan usually improves the quality, so a viewer can show a rough picture first and then make it sharper.
  • Lossless does not use the DCT. The decoded picture is identical to the original.
  • differential belongs to the hierarchical mode. Here each later frame stores only the differences from the frame before.
  • Huffman coding and arithmetic coding are 2 lossless ways to pack the data into fewer bits.

Why it matters for your file

The JPEG standard does not require every program to support every process. Every DCT decoder must support baseline, so a baseline file is the safe choice when a picture must open everywhere.

Other processes can fail to open in some programs. According to the Library of Congress, the original lossless process was not widely supported in applications. One lossless variant with Huffman coding is used in version 1.1 of the Adobe DNG raw format.

A value with DCT tells you that some detail was dropped when the picture was saved. Saving the file again does not bring that detail back. The field only describes how the picture is stored and holds no personal information.