Clarify 8 bit in Extended JPEG Transfer Syntax

DICOM provides a mechanism for supporting the use of JPEG Image Compression through the Encapsulated Format (see ). defines a number of Transfer Syntaxes that reference the JPEG Standard and provide a number of lossless (bit preserving) and lossy compression schemes. The context where the usage of lossy compression of medical images is clinically acceptable is beyond the scope of the DICOM Standard. The policies associated with the selection of appropriate compression parameters (e.g., compression ratio) for JPEG lossy compression is also beyond the scope of this standard. In order to facilitate interoperability of implementations conforming to the DICOM Standard that elect to use one or more of the Transfer Syntaxes for JPEG Image Compression, the following policy is specified: Any implementation that conforms to the DICOM Standard and has elected to support any one of the Transfer Syntaxes for lossless JPEG Image Compression, shall support the following lossless compression: The subset (first-order horizontal prediction [Selection Value 1) of JPEG Process 14 (DPCM, non-hierarchical with Huffman coding) (see ). Any implementation that conforms to the DICOM Standard and has elected to support any one of the Transfer Syntaxes for 8-bit lossy JPEG Image Compression, shall support the JPEG Baseline Compression (coding Process 1). Any implementation that conforms to the DICOM Standard and has elected to support any one of the Transfer Syntaxes for 12-bit lossy JPEG Image Compression, shall support the JPEG Compression Process 4. ...

DICOM defines defaults for Lossy JPEG Image Compression, one for 8-bit images and the other for 12-bit images. JPEG coding Process 1 (identified by Transfer Syntax UID = "1.2.840.10008.1.2.4.50") is used for 8-bit images. JPEG coding Process 4 (identified by Transfer Syntax UID = "1.2.840.10008.1.2.4.51") is used for 12-bit images. This implies that: If an Application Entity issues an A-ASSOCIATE request where any offered Abstract Syntaxes is associated in one or more Presentation Context(s) with a JPEG lossy compression Transfer Syntax, at least one of the Presentation Contexts that include this Abstract Syntax, shall include the appropriate DICOM Default Lossy JPEG Compression Transfer Syntax. Offering Abstract Syntax (AS1) in two Presentation Contexts with Transfer Syntaxes JPEG lossy (JL1) and (JL2) is not valid, but offering AS1-JL1, AS1-JL2 and AS1-JLD is valid because the DICOM Default JPEG Lossy Transfer Syntax (JLD) is present in at least one of the Presentation Contexts that are based on Abstract Syntax (AS1).2. The DICOM Default Transfer Syntax (uncompressed) may be offered if the sender has access to the original pixel data in an uncompressed or lossless compressed form. If an Application Entity that supports one or more Lossy JPEG Transfer Syntaxes receives an A-ASSOCIATE indication corresponding to a request that follows the requirements specified in (a), every Presentation Context related to a given Abstract Syntax cannot be rejected in an A-ASSOCIATE response for the reason that the DICOM Default lossy JPEG Transfer Syntax is not supported. The 12 bit default Transfer Syntax 1.2.840.10008.1.2.4.51 can also be used to encode 8 bit images, but the bit stream required is not identical to that used in the 8 bit default Transfer Syntax 1.2.840.10008.1.2.4.50 (see ).

The International Standards Organization ISO/IEC JTC1 has developed an International Standard, ISO/IS- 10918-1 (JPEG Part 1) and an International Draft Standard, ISO/IS- 10918-2 (JPEG Part 2), known as the JPEG Standard, for digital compression and coding of continuous-tone still images (see for further details). A DICOM Transfer Syntax for JPEG Image Compression shall be identified by a UID value, appropriate to its JPEG coding process, chosen from .

DICOM Transfer Syntax UIDs for JPEG

DICOM Transfer Syntax UID	JPEG coding process	JPEG description
1.2.840.10008.1.2.4.50	1	baseline
1.2.840.10008.1.2.4.51	2(8-bit),4(12-bit)	extended
1.2.840.10008.1.2.4.57	14	lossless, non-hierarchical
1.2.840.10008.1.2.4.70	14 (Selection Value 1)	lossless, non-hierarchical, first-order prediction

DICOM identifies, to increase the likelihood of successful association, three Transfer Syntaxes for Default JPEG Compression Image processes (see and ). Different JPEG processes may use different SOF marker segments. E.g., the baseline JPEG process 1 used with the 1.2.840.10008.1.2.4.50 Transfer Syntax uses the SOF0 marker, whereas the extended process 2 used with the 1.2.840.10008.1.2.4.51 Transfer Syntax uses the SOF1 marker. Accordingly, even though both bit streams encode 8 bit images using DCT and Huffman coding, they are not identical. Further, the extended process 2 may (but is not required to) use more AC and DC tables (up to 4 of each, rather than 2, per ISO 10918-1 Section F.1.3). It is not compliant to send bit streams with the SOF0 marker using the 1.2.840.10008.1.2.4.51 Transfer Syntax, but it is recommended that receivers of the 1.2.840.10008.1.2.4.51 Transfer Syntax be able to decode bit streams with the SOF0 marker (this asymmetry is consistent with ISO 10918-2 requirements; see ). It is recommended that lossy compressed 8 bit images be encoded with the 1.2.840.10008.1.2.4.50 Transfer Syntax rather than the 1.2.840.10008.1.2.4.51 Transfer Syntax, unless the additional features of the extended process are required, particularly since support of the 1.2.840.10008.1.2.4.50 Transfer Syntax is required for 8 bit images anyway (as described in ) and to avoid confusion with the use of 12 bit images encoded with Process 14 in the 1.2.840.10008.1.2.4.51 Transfer Syntax (defined as a DICOM Default Transfer Syntax for 12 bit images in ). If the object allows multi-frame images in the pixel data field, then each frame shall be encoded separately. Each fragment shall contain encoded data from a single-frame image. Though a fragment may not contain encoded data from more than one frame, the encoded data from one frame may span multiple fragments. See note in . For all images, including all frames of a multi-frame image, the JPEG Interchange Format shall be used (the table specification shall be included). This refers to the ISO 10918-1 "interchange format", not the DISISO 10918-5 JPEG File Interchange Format (JFIF). If images with Photometric Interpretation (0028,0004) YBR_FULL_422 or YBR_PARTIAL_422, are encoded with JPEG coding Process 1 (non hierarchical with Huffman coding), identified by DICOM Transfer Syntax UID "1.2.840.10008.1.2.4.50"the minimum compressible unit is YYCBCR, where Y, CB, and CR are 8 by 8 blocks of pixel values. The data stream encodes two Y blocks followed by the corresponding CB and CR blocks.

The International Standards Organization (ISO/IEC JTC1/SC2/WG10) has prepared an International Standard, ISO/IS- 10918-1 (JPEG Part 1) and International Draft Standard ISO/IS- 10918-2 (JPEG Part 2), for the digital compression and coding of continuous-tone still images. This standard is collectively known as the JPEG Standard. Part 1 of the JPEG Standard sets out requirements and implementation guidelines for the coded representation of compressed image data to be interchanged between applications. The processes and representations are intended to be generic in order to support the broad range of applications for color and grayscale still images for the purpose of communications and storage within computer systems. Part 2 of the JPEG Standard defines tests for determining whether implementations comply with the requirements of the various encoding and decoding processes specified in Part 1 of the JPEG Standard. The JPEG Standard specifies lossy and lossless code processes. The lossy coding is based on the discrete cosine transform (DCT), permitting data compression with an adjustable compression ratio. The lossless coding employs differential pulse code modulation (DPCM). The JPEG Standard permits a variety of coding processes for the coder and decoder. These processes differ in coding schemes for the quantified data and in sample precision. The coding processes are consecutively numbered as defined in the International Draft Standard ISO/IS- 10918-2 (JPEG Part 2), and are summarized in . The simplest DCT-based coding process is referred to as Baseline Sequential with Huffman Coding for 8-bit Samples.

JPEG Modes of Image Coding

No.	Description	Lossy LY Lossless LL	Non-Hierarchical NH Hierarchical H	Sequential S Progressive P	Transform	Coding	Accepted Bits
1 2 4	Baseline Extended Extended	LY LY LY	NH NH NH	S S S	DCT DCT DCT	Huffman Huffman Huffman	8 8 12
14	Lossless	LL	NH	S	DPCM	Huffman	2-16

The different coding processes specified in the JPEG Standard are closely related. By extending the capability of an implementation, increasingly more 'lower level' processes can also be executed by the implementation. This is shown in for Huffman Coding. Inclusion of a JPEG-coded image in a DICOM message is facilitated by the use of specific Transfer Syntaxes that are defined in . Independent of the JPEG coding processes, the same syntax applies. The only distinction for different processes in the syntax (apart from different SOF marker segments in the JPEG bit stream) is the UID value. lists the UID values in the Transfer Syntax for the various JPEG coding processes for reference.

Relationship Between the Lossy JPEG Huffman Coding Processes

Process	1	2	4
1	*	*	*
2		*	*
4			*

* Coding process of column can execute coding process of row

Identification of JPEG Coding Processes in DICOM

DICOM Transfer Syntax UID	JPEG process	JPEG description	capable of performingdecoding
1.2.840.10008.1.2.4.50	1	baseline	1
1.2.840.10008.1.2.4.51	2,4	extended	1,2,4 (see Note)
1.2.840.10008.1.2.4.57	14	lossless NH	14
1.2.840.10008.1.2.4.70	14 Selection Value 1	lossless NH, first-order prediction

Though the coding processes (2, 4) described in ISO 10918-1 are capable of decoding the other listed process (1), the bit stream uses different SOF marker segments. I.e., the baseline JPEG process 1 used with the 1.2.840.10008.1.2.4.50 Transfer Syntax uses the SOF0 marker, whereas the extended process 2 used with the 1.2.840.10008.1.2.4.51 Transfer Syntax uses the SOF1 marker. Accordingly, even though both bit streams encode 8 bit images using DCT and Huffman coding, they are not identical. ISO 10918-2 describes compliance tests for decoders, and requires that implementations of specific extended processes (such as 2 and 4) be capable of decoding bit streams of related baseline processes (such as 1) (ISO 10918-2 Section 7.4 Compliance tests for DCT-based sequential mode decoding processes). The converse is not true for encoders however, and the presence of SOF marker segments not defined by the specific process is not compliant (ISO 10918-2 Section 5.1.1 Non-hierarchical coding processes syntax compliance test Tables 1 and 2).