JPEG XR Codec Overview

• 05/31/2018
• 13 minutes to read
• Contributors
  • 
  • 

The native JPEG XR codec is available through the Windows Imaging Component (WIC). The JPEG XR format, which the codec supports, is designed for consumer and professional digital photography.

The JPEG XR format can achieve up to twice the compression efficiency of the original JPEG format, with less noticeable compression artifacts. Features of JPEG XR include:

• Support for monochrome, RGB, CMYK, and n-channel images
• 8-, 16-, and 32-bit integer formats
• High dynamic range, wide-gamut formats, using fixed point or floating point color values
• Progressive decoding
• Lossy or lossless encoding using the same compression algorithm
• Support for decoding of regions of interest in large images

The JPEG XR format is defined in the following standards documents:

• ITU-T T.832: Information technology – JPEG XR image coding system – Image coding specification
• ISO/IEC 29199-2:2010: Information technology — JPEG XR image coding system — Part 2: Image coding specification

The JPEG XR standard is largely based on the HD Photo format, but there are some differences between the two formats. In Windows 8, the HD Photo codec has been updated to support JPEG XR. The encoder now always outputs a JPEG XR–compliant bit stream. The decoder can decode both JPEG XR and HD Photo images.

Substantial performance improvements, in relation to the HD Photo codec, have been made to the JPEG XR codec. For example, sub-resolution image decoding, such as thumbnail generation, has been improved, as well as low-resolution image decoding. It is recommended that you use the JPEG XR format instead of the HD Photo format.

Codec Information

Codec Features

High Dynamic Range

JPEG XR supports high-dynamic range images, using floating-point or fixed-point colors. In these color formats, the numerical range of a pixel is larger than the visible range, so you can adjust colors above or below the visible range during intermediate processing stages.

• Fixed-point: In a fixed-point representation, 0 represents black and 1.0 represents maximum saturation. The JPEG XR codec supports both 16-bit and 32-bit fixed-point formats. For 16-bit, 1.0 = 0x2000h, which gives 13 bits for the visible range [0...1]. The total range is –4.0 to +3.999, and is mapped linearly. For 32-bit, 1.0 = 0x01000000h, the visible range is 24 bits, and the total range is –128 to +127.999.
• Floating-point: In a floating-point representation, 0 represents black and 1.0 represents maximum saturation. The JPEG XR codec supports both 16-bit and 32-bit floating-point formats.

Tiles

A frame can be partitioned into rectangular subregions called tiles. A tile is an area of a image that contains rectangular arrays of macroblocks. Tiles enable regions of the image to be decoded without processing the entire image.

During encoding, select the number of tiles by setting the HorizontalTileSlices and VerticalTileSlices properties. The minimum tile size is 16 × 16 pixels. The encoder adjusts the number of tiles to maintain this restriction. There is storage and processing overhead associated with each tile, so you should consider the number of tiles that are needed for particular scenarios.

Image Stream Output

The JPEG-XR standard defines two parts of a JPEG-XR file:

• The image bit stream, defined in the body of the standard.
• The image container. The file contains Exif and XMP metadata, and is defined in Annex A of the standard.

It is possible, and allowed by the standard, to embed the image stream inside another type of file container. The encoder supports a stream-only mode, which outputs the raw image bit stream with no image container. An application can store the bit stream in some other container format.

To enable stream-only mode, set the StreamOnly property.

Image Quality Settings

Several codec properties control the quality of the output image from the encoder.

• ImageQuality is a property common across WIC codecs. It specifies image quality as a single floating-point value from 0.0–1.0,
• The Quality, Overlap, and Subsampling properties give more control over the quality settings.

To use the Quality, Overlap, and Subsampling properties, set the UseCodecOptions property to VARIANT_TRUE.

If UseCodecOptions is VARIANT_FALSE (VARIANT_FALSE is the default) the encoder uses the ImageQuality property. The encoder maps the value of ImageQuality to Quality, Overlap, and Subsampling through a lookup table.

The encoder does not support the CompressionQuality property.

Compressed Domain Transcode

The JPEG XR codec can perform certain image transformations without actually decoding the compressed data and re-encoding it. Compressed domain operations are very efficient and avoid any additional quality loss that is typical when you decode and re-encode a lossy-compressed image.

The following compressed domain operations are supported:

• Crop a region of the image.
• Rotate or flip the image.
• Discard frequency data to create a smaller image file.
• Reorganize the image between spatial and frequency order.

The JPEG XR encoder uses compressed domain transcoding, if possible, when the source image is a JPEG XR image. When the encoder performs a compressed domain operation, it ignores the following codec properties: AlphaQuality, ImageQuality, InterleavedAlpha, LosslessOverlap, and Quality. If the HorizontalTileSlices and VerticalTileSlices properties are present, you must set them to zero. You cannot change the tile size of an image as part of a compressed domain transcode.

The follow list describes how to perform the image transformations.

• To crop the image, set the desired region in the WICRect parameter of the WriteSource method.
• To rotate or flip the image, set the BitmapTransform property.
• To discard frequency data in the image, set the ImageDataDiscard property. To discard frequency data in the alpha channel, set the AlphaDataDiscard property. Discarding frequency data reduces the encoded file size and can reduce the resolution.
• To change the image organization between frequency and spatial ordering, set the FrequencyOrdering property.

To disable compressed domain transcode and force the encoder to re-encode the image, set the UseCodecOptions to VARIANT_TRUE and set CompressedDomainTranscode to VARIANT_FALSE.

Encoder Options

To set encoding properties, use the IPropertyBag2 interface. For more information, see the Encoding Overview.

The following list specifies the encoder options.

• AlphaDataDiscard
• AlphaQuality
• BitmapTransform
• CompressedDomainTranscode
• FrequencyOrder
• HorizontalTileSlices
• IgnoreOverlap
• ImageDataDiscard
• ImageQuality
• InterleavedAlpha
• Lossless
• Overlap
• ProgressiveMode
• Quality
• StreamOnly
• Subsampling
• UseCodecOptions
• VerticalTileSlices

AlphaDataDiscard

Sets the amount of alpha frequency data to discard during a compressed domain transcode.

This property applies only if the CompressedDomainTranscode property is set to VARIANT_TRUE and the image contains either a planar alpha channel or interleaved alpha channel; otherwise it is ignored.

For images that contain a planar alpha channel, the following values are valid.

For images that contain an interleaved alpha channel, the following value is valid.

For interleaved alpha, unless this property is set to 4, the alpha channel is processed the same as the image data, according to the value of the ImageDataDiscard property.

AlphaQuality

Sets the compression quality for the planar alpha channel image.

This property applies when the image has an alpha channel and the InterleavedAlpha property is VARIANT_FALSE. The value 1 indicates lossless mode. Increasing values result in higher compression ratios and lower image quality.

BitmapTransform

Specifies whether the image is rotated or flipped when decoded.

CompressedDomainTranscode

Enables or disables compressed domain transcoding.

To disable compressed domain operations, set this property to VARIANT_FALSE.

FrequencyOrder

Enables encoding in frequency order. Device implementations of JPEG XR encoders can organize a file in spatial order to reduce the memory required during encoding.

• VARIANT_TRUE: Frequency order. The lowest frequency data appears first in the file, and image content is grouped by its frequency rather than its spatial orientation. Organizing a file by frequency order provides the best performance for any frequency-based decoding.
• VARIANT_FALSE: Spatial order. Spatial order reduces the memory required during encoding

Frequency order is recommended unless you have performance or application-specific reasons to use spatial order.

HorizontalTileSlices

Sets the number of horizontal tiles.

The value is the number of horizontal subdivisions; that is, the number of horizontal tiles – 1.

IgnoreOverlap

Specifies how the encoder handles tile boundaries during a compressed domain transcode.

This property is applied only if the CompressedDomainTranscode property is set to VARIANT_TRUE and a sub-region transcode of exactly one or more tiles is performed.

The default operation for a region transcode is to expand the requested region to include the surrounding pixels that are required for overlap decoding of the region edges. If this property is set to VARIANT_TRUE, the codec ignores the surrounding pixels, and only the selected tile or tiles are extracted. If the source image is not tiled or if the requested region includes partial tiles, this parameter is ignored.

ImageDataDiscard

Sets the amount of image frequency data to discard during a compressed domain transcode.

This property applies only if the CompressedDomainTranscode property is set to VARIANT_TRUE; otherwise it is ignored.

If the image contains an interleaved alpha channel, the value of ImageDataDiscard is applied to the alpha channel unless the AlphaDataDiscard property is set to 4, in which case the alpha channel is discarded.

For planar alpha, the frequency data that is discarded is controlled by the AlphaDataDiscard property.

ImageQuality

Sets the image quality.

Level 1.0 gives mathematically lossless compression.

Level 0.0 is the lowest quality setting.

InterleavedAlpha

Specifies whether to encode interleaved alpha or planar alpha.

• VARIANT_TRUE: Interleaved alpha. Alpha channel information is encoded as an additional interleaved channel, with no correlation to the image content channels. This mode is useful for decoding alpha simultaneously with the image when the image is streaming.
• VARIANT_FALSE: Planar alpha. A planar alpha channel is encoded as a separate image. The image data and the alpha channel are decoded independently. Optionally, you can set the quality level of the alpha channel by setting the AlphaQuality property.

Interleaved alpha is supported only for certain RGB pixel formats. Planar alpha is supported for any image format that defines an alpha channel.

Lossless

Enables losses compression.

If the value is VARIANT_TRUE, the encoder uses lossless compression. When set to VARIANT_TRUE, this property overrides the ImageQuality property.

Overlap

Sets the level of overlap filtering. With overlap filtering, transform coefficients are applied across block and macroblock boundaries. This can reduce blocking artifacts.

Definitions:

• One level of overlap: The encoded values of 4x4 blocks are modified based on neighboring blocks.
• Two levels of overlap: The first level of overlap is applied. In addition, the encoded values of 16x16 macroblocks are modified based on the neighboring macroblocks.
• Soft tiling: Overlap filtering is applied across tile boundaries.
• Hard tiling: Overlap filtering is not applied across tile boundaries. Hard tiles can introduce some visual artifacts along tile boundaries.

If you set this property, also set UseCodecOptions to VARIANT_TRUE.

ProgressiveMode

Enables or disables progressive encoding.

Quality

Sets the compression quality.

The value 1 indicates lossless mode. Increasing values result in higher compression ratios and lower image quality.

If you set this property, also set UseCodecOptions to VARIANT_TRUE.

StreamOnly

Enables or disables stream-only mode.

Subsampling

Sets the chroma subsampling. This property applies only to RGB images.

4:2:2 and 4:2:0 preserve luminance detail at the expense of color detail.

If you set this property, also set UseCodecOptions to VARIANT_TRUE.

UseCodecOptions

Specifies whether to use the Quality, Overlap, and Subsampling properties instead of the generic ImageQuality property.

VerticalTileSlices

Sets the number of horizontal tiles.

The value is the number of vertical subdivisions; that is, the number of vertical tiles – 1.

Supported Color Formats

For more information about these formats, see Native Pixel Formats.

• Integer RGB formats
  • GUID_WICPixelFormat24bppRGB
  • GUID_WICPixelFormat24bppBGR
  • GUID_WICPixelFormat32bppBGR
  • GUID_WICPixelFormat48bppRGB
  • GUID_WICPixelFormat32bppBGRA
  • GUID_WICPixelFormat64bppRGBA
  • GUID_WICPixelFormat32bppPBGRA
  • GUID_WICPixelFormat64bppPRGBA
• Fixed-point RGB formats
  • GUID_WICPixelFormat48bppRGBFixedPoint
  • GUID_WICPixelFormat64bppRGBFixedPoint
  • GUID_WICPixelFormat96bppRGBFixedPoint
  • GUID_WICPixelFormat128bppRGBFixedPoint
  • GUID_WICPixelFormat128bppRGBAFixedPoint
• Floating-point RGB formats
  • GUID_WICPixelFormat48bppRGBHalf
  • GUID_WICPixelFormat64bppRGBHalf
  • GUID_WICPixelFormat128bppRGBFloat
  • GUID_WICPixelFormat64bppRGBAFixedPoint
  • GUID_WICPixelFormat64bppRGBAHalf
  • GUID_WICPixelFormat128bppRGBAFloat
  • GUID_WICPixelFormat128bppPRGBAFloat
• Grayscale formats
  • GUID_WICPixelFormat8bppGray
  • GUID_WICPixelFormat16bppGray
  • GUID_WICPixelFormat16bppGrayFixedPoint
  • GUID_WICPixelFormat16bppGrayHalf
  • GUID_WICPixelFormat32bppGrayFixedPoint
  • GUID_WICPixelFormat32bppGrayFloat
• Packed formats
  • GUID_WICPixelFormat16bppBGR555
  • GUID_WICPixelFormat16bppBGR565
  • GUID_WICPixelFormat32bppBGR101010
  • GUID_WICPixelFormat32bppRGBE
• CMYK formats
  • GUID_WICPixelFormat40bppCMYKAlpha
  • GUID_WICPixelFormat64bppCMYK
  • GUID_WICPixelFormat80bppCMYKAlpha
• N-channel formats
  • GUID_WICPixelFormat32bpp4Channels
  • GUID_WICPixelFormat40bpp5Channels
  • GUID_WICPixelFormat48bpp6Channels
  • GUID_WICPixelFormat56bpp7Channels
  • GUID_WICPixelFormat64bpp8Channels
  • GUID_WICPixelFormat32bpp3ChannelsAlpha
  • GUID_WICPixelFormat40bpp4ChannelsAlpha
  • GUID_WICPixelFormat48bpp5ChannelsAlpha
  • GUID_WICPixelFormat56bpp6ChannelsAlpha
  • GUID_WICPixelFormat64bpp7ChannelsAlpha
  • GUID_WICPixelFormat72bpp8ChannelsAlpha
  • GUID_WICPixelFormat48bpp3Channels
  • GUID_WICPixelFormat64bpp4Channels
  • GUID_WICPixelFormat80bpp5Channels
  • GUID_WICPixelFormat96bpp6Channels
  • GUID_WICPixelFormat128bpp8Channels
  • GUID_WICPixelFormat64bpp3ChannelsAlpha
  • GUID_WICPixelFormat80bpp4ChannelsAlpha
  • GUID_WICPixelFormat96bpp5ChannelsAlpha
  • GUID_WICPixelFormat112bpp6ChannelsAlpha
  • GUID_WICPixelFormat128bpp7ChannelsAlpha
  • GUID_WICPixelFormat144bpp8ChannelsAlpha