JPEG-LS

The Joint Photographic Experts Group, in addition to the well-known JPEG and JPEG 2000 image compression formats, focused primarily on lossy compression, also proposed a lossless compression standard, JPEG-LS (which, however, also provides limited loss compression mode).

Content

JPEG-LS format

The JPEG-LS format was based on the LOCO-I (Low Complexity Lossless Compression for Images) format ^[1] . The lossless compression algorithm LOCO-I, adopted as the basis for the development of the JPEG-LS standard, provided for the first time not only lossless , but also near lossless mode (compression with limited, user-defined losses). The JPEG-LS decoder is almost the same as the encoder, so this compression algorithm is symmetrical. [one]

The compression algorithm underlying JPEG-LS uses an adaptive prediction of the current pixel value by the environment including already encoded pixels (Median Edge Detection method), context classification, context modeling of the prediction error and its correction, as well as the entropy coding of the corrected prediction error ( Golomb- Rice coding) ^[2] . To improve the coding efficiency of low-entropy images (or image fragments), the algorithm provides for automatic transition to the run -length coding mode, which allows it to be used for lossless compression (or with limited losses) of not only photorealistic images, but also computer graphics.

For color (multi-component) images, the JPEG-LS standard does not prescribe any particular method for converting color components, so programs implementing JPEG-LS usually suggest choosing one scheme from several (independent compression of color planes, JPEG 2000 lossless mode conversion and etc.)

Application

The JPEG-LS format was developed primarily for storing images for medical purposes, that is, for those cases where it is important to have a large image without any loss of quality. As already mentioned, the LOCO-I format, developed in the walls of HP Labs, was taken as the basis. It was then finalized by the joint efforts of Hewlett-Packard and Mitsubishi . Both companies allowed to use their patents for this format without paying for a license, so JPEG-LS can also be found in regular PC programs. [2]

Competitive Formats

Lossless JPEG

Lossless JPEG is an addition to JPEG (and to JPEG-LS, despite some similarities in the title, it has no relation). Unlike “regular” JPEG, built on the basis of discrete cosine transform, Lossless JPEG uses a pixel prediction scheme for the three nearest neighbors — the upper, left, and upper left pixels — and uses entropy coding to compress the difference between the true and predicted pixel values. Unlike JPEG-LS, the Lossless JPEG compression algorithm provides neither adaptive prediction of the value of the encoded pixel, nor contextual modeling of the prediction error. Lossless JPEG uses a Huffman code to entropy encode prediction errors. As an alternative, the standard allows the use of arithmetic coding , however, due to patent restrictions, it has not found application in practical implementations of Lossless JPEG. This method is not widely used and is not supported by the popular IJG libraries .

JPEG 2000

JPEG 2000 also contains a lossless compression mode (different from JPEG-LS) based on a special integer wavelet filter (3/5 biorthogonal). Lossless compression in JPEG 2000 is slower and produces slightly less compression than JPEG-LS on both artificial and photorealistic images [3] [4] [5] [6] .

Notes

Links

[1] MJ Weinberger, G. Seroussi, G. Sapiro "The LOCO-I Lossless Image Compression Algorithm: Principles and Standardization into JPEG-LS."

[2] SW Golomb Run-length encodings.