ImageGear Supported Versions:

• Version 1.01 - 1991
• Version 1.02 - Added ability for thumbnails to be color-mapped and JPEG compressed.

ImageGear Supported Features:

• IG_FLTR_DETECTSUPPORT - autodetection
• IG_FLTR_PAGEREADSUPPORT - single page file reading
• IG_FLTR_PAGEINSERTSUPPORT - single-page file writing

ImageGear Read Support:

• IG_COMPRESSION_JPEG (Lossy):
  • Grayscale: 8, 12 bpp;
  • RGB: 24, 36 bpp;
  • CMYK: 32 bpp;
  • RGB + Alpha: 32 bpp
• IG_COMPRESSION_JPEG (Lossless):
  • Grayscale: 8, 16 bpp;
  • RGB: 24, 48 bpp;
  • CMYK: 32 bpp;
  • RGB + Alpha: 32 bpp
• IG_COMPRESSION_JPEG (Progressive JPEG):
  • Grayscale: 8, 12 bpp;
  • RGB: 24 bpp;
  • CMYK: 32 bpp;
  • RGB + Alpha: 32 bpp

ImageGear Write Support:

• IG_COMPRESSION_JPEG (Lossy):
  • Grayscale: 8, 12 bpp;
  • RGB: 24, 36 bpp;
  • RGB + Alpha: 32 bpp1
  • CMYK: 32 bpp2
• IG_COMPRESSION_JPEG (Lossless):
  • Grayscale: 8, 16 bpp;
  • RGB: 24 bpp;
• IG_COMPRESSION_JPEG (Progressive JPEG):
  • Grayscale: 8, 12 bpp;
  • RGB: 24, 36 bpp

1. RGBA saving is disabled by default. Set SAVE_ALLOW_RGBA control parameter to TRUE to enable it.
2. CMYK saving is enabled by default. You can disable it by setting SAVE_ALLOW_CMYK control parameter to FALSE.

ImageGear Filter Control Parameters:

Comments:

JPEG is normally associated with the JPEG compression scheme, but it is also implemented into the JFIF file format. This format was developed to store JPEG-encoded data, and to exchange it between applications or operating systems that are normally incompatible.

The header of the JFIF contains the version number, the image dots per inch (DPI), or dots per centimeter, and an optional thumbnail (miniature) RGB representation of the main image. Version 1.02 handles thumbnails differently by storing them separately, rather than in the identification marker of the header.

The raw JPEG data is surrounded by two markers, an "SOI" (start of image) marker, and an "EOI" (end of image) marker. See the section entitled Compression Schemes for more about JPEG compression.

JFIF is considered a non-proprietary file format. Many proprietary file formats contain JPEG data, incorporating their own application-specific data structures. Other non-proprietary formats that use JPEG-encoded data: Macintosh MAC PICT and TIFF file format, version 6.0.

Loading and Saving JPEG-Compressed Images

ImageGear supports the reading and writing of three types of JPEG compression: baseline JPEG (Lossy), Progressive JPEG, and Lossless JPEG. When you load a JPEG-compressed file, ImageGear detects the type of JPEG compression and decompress the image automatically. But if you want to save an image with a JPEG compression scheme other than baseline JPEG, you must use the "SAVE_TYPE" parameter to specify the type of JPEG compression.

Use these constants for "SAVE_TYPE" control parameter:

• IG_JPG_LOSSY- Lossy JPEG compression.
• IG_JPG_LOSSLESS - Lossless JPEG compression
• IG_JPG_PROGRESSIVE - Progressive JPEG compression.

JPEG Decimation

This table lists all possible decimation values:

The format of these ImageGear decimation constants is:

IG_JPG_DCM_<H1>x<V1>_<H2>x<V2>_<H3>x<V3>,

where Hi, Vi = horizontal and vertical decimation values for the i-channel.

The following is a simple example of decimation. For a more detailed definition, please see the JPEG Specification.

A decimation setting of IG_JPG_DCM_4x2_1x1_1x1 would yield the following results:

As shown, 8 Y components in the source image have yielded one Cb and one Cr component. In general, this setting will reduce the quality of the compression, unless the image has many continuous tone areas.

• Maximum quality can be reached using a value of IG_JPG_DCM_1x1_1x1_1x1.
• Maximum compression ratio can be reached with a value of IG_JPG_DCM_4x4_2x2_2x2.
• The ImageGear default decimation value is: IG_JPG_DCM_2x2_1x1_1x1.

JPEG Thumbnails

The JPEG format can store thumbnails, which are small representations of the original image. These images are stored in uncompressed form and can significantly decrease your overall compression ratio. (Uncompressed thumbnails sometimes occupy more space than the original JPEG image when compressed). Use this option carefully.

Lossless JPEG Predictor Settings

The Lossless JPEG scheme is "predictive" in nature-it uses the values of surrounding pixels in addition to the value of the original pixel to calculate a predictor value, which it then subtracts from the value of the original pixel. The resulting pixel value will be reduced such that it can be compressed more than the original value. The higher the number of neighboring pixels used, the higher the compression will be.

Regardless of the predictor value setting, the quality of the image will remain the same. The difference is that if you choose to optimize for space by setting a high predictor value, you will have to give up some speed, as the decompression will take longer to perform.

ImageGear lets you set the predictor value using "PREDICTOR" control parameter. The ImageGear default for this setting is 1. The allowed range is 1-7. Values outside of the allowed range of 1-7 will be mapped to a value of 5. The graphic below shows a predictor (x) and three reconstructed samples (a,b,c) immediately to the left, immediately above, and diagonally to the left of the current sample

Lossless JPEG does not apply DCT for an image as per the Lossy JPEG compression. Instead, it uses a DPCM difference coding, which can be carried out with any one of seven different prediction modes. Correspondingly, the IG_CONTROL_JPG_PREDICTOR control parameter can be set to a value between 1 and 7. In Table 4, you will see what algorithm your setting of 1-7 will use, where Px is the predictor, and Ra, Rb, and Rc are the reconstructed samples:

JPEG Quality

The baseline JPEG specification calls for a quality setting. The lower the setting, the greater the number of original pixels lost, and therefore the smaller the resulting compressed file will be. ImageGear lets you set the quality of compression with values of type INT between 1-100, where 100 provides the highest retention of original pixel values. A setting of 100 does not mean that the image includes 100% of all original pixel values. With Lossy JPEG, there is no such thing as "no loss". Control parameter for setting Lossy JPEG quality is "QUALITY".

JPEG Entropy Optimization

Entropy optimization is relevant only to standard Lossy JPEG compression. If you set the parameter to TRUE, the default Huffman tables are not used. Instead, optimal Huffman tables are created for each component. This can bring a higher compression ratio but it takes more time for the compression.

Progressive JPEG Scans

A Progressive JPEG file stores more than just a copy of an image but rather several scans, each of which progressively adds a higher level of quality. Each scan contains a portion of the original image data. The purpose of this is to allow a very quick display of an image, beginning with a low-quality rendering and then increasing in quality as the remaining scans are added to it.

A Progressive JPEG image is stored as sequence of Huffman compressed blocks or "scans". Each scan contains the sequence of DCT coefficients in the given range. However, the coefficients are not complete. Only some of their bits will be stored in each scan.

ImageGear defines the following structure for holding the necessary configuration to write a JPEG image:

typedef struct tag AT_PJPEG_SCANINFO{
                LONG            Ss;
                LONG            Se;
                LONG            HBit;
                LONG            LBit;
                LONG            ChannelID;
}               AT_PJPEG_SCANINFO;
typedef AT_PJPEG_SCANINFO FAR* LPAT_PJPEG_SCANINFO;

The Ss and Se members of the AT_PJPEG_SCANINFO structure are used for spectral selection control coefficients:

LONG                 Ss; 
LONG                    Se; - after applying DCT of 8x8 pixels we get 64 

In Progressive coding, these coefficients are separated into different scans. Values Ss and Se specify the first and last number of the DCT coefficients that must be included in a given scan. The possible values are Ss = Se = 0 or 1<= Ss<=Se<=63. Please note following restrictions:

• The first coefficient (DC) cannot be encoded with the other coefficient (AC) in the single scan. In other words, the DC and AC coefficients cannot be in the same scan.
• Only scans that code DC coefficients may include interleaved blocks from more than one component. All other scans shall have only one component. For each component, a first DC scan shall precede any AC scans.

The HBit and LBit members of the AT_PJPEG_SCANINFO structure are used for successive approximation control:

        LONG HBit;
        LONG LBit;

If successive approximation is used, the DCT coefficients are reduced in precision by the point transform defined in the scan header. This is equivalent to taking some binary digits from each coefficient. HBit and LBit specify the high and low range of bits to take. For example, if HBit = 7 and LBit = 2, the scan will have the following original bits of the original DCT coefficient: 7,6,5,4,3,2.

The ChannelID member of the AT_PJPEG_SCANINFO structure is used to specify the number of components that will be encoded:

LONG ChannelID;

It can be set to one of the following values:

• interleaved scan which will only have DC coefficients of all components. This setting can only be used if Ss=Se=0.
• takes coefficients of first component.
• takes coefficients of second component.
• takes coefficients of third component.

There are two control parameters that operate with the AT_PJPEG_SCANINFO structure:

The LOAD_SCANS control parameter specifies how many scans should be loaded. For example, if it is set to 1, the JPEG filter will load only the first scan of the image.

The following text blocks represent the different AT_PJPEG_SCANINFO structures that would be generated when loading a JPEG file using the ImageGear default settings for Progressive scans:

References Used:

Brown, C. Wayne, and Barry J. Shepherd. Graphics File Formats: Reference and Guide. Greenwich, CT.: Manning Publications, 1992.

Kay, David C. and John R. Levine. Graphics File Formats. Windcrest Books, 1992.

Murray, James D. and William vanRyper. Encyclopedia of Graphics File Formats. Sebastopol, CA: O'Reilly & Associates, Inc., 1994.