Digital Negative Specification 1.7.0.0 June 2023
16
Depth information is stored as an 8 or 16-bit single channel unsigned integer image. Zero encodes the
“nearest distance” (closest to the camera), and the maximum possible unsigned integer encodes the
“farthest distance”. The mapping of the depth map integer values to distance values is specified by the
DepthFormat, DepthNear, DepthFar, DepthUnits, and DepthMeasureType tags.
The depth map should match the field of view of the main image. It may have a different pixel resolution
than the main image, but if the depth map is resized to match the size of the main image, the pixels should
correspond. The depth values should be filled in by an estimation algorithm for any areas of the depth map
that cannot be measured directly (i.e. no holes are allowed in depth maps).
Enhanced Image Data
DNG 1.5.0.0 adds the concept of enhanced image data, which is a more processed (and always at least
demosaiced) version of the raw image data.
Enhanced image data is stored in an IFD with a NewSubFileType of 16 (see section on NewSubFileType for
details).
The enhanced image data should always use a PhotometricInterpretation of 34892 (LinearRaw). It should use
the same color space as the raw image data.
The enhanced image data should use either 16-bit unsigned integer encoding, or 16 or 24 or 32-bit floating
point encoding.
The enhanced image data is assumed to have already been mapped to linear reference values, with the
exception that it may have a non-zero BlackLevel tag if unsigned integer encoding is used. This black level is
required to be the same for all color planes of the enhanced image data.
The enhanced IFD may have its own copies of the BaselineSharpness, NoiseProfile, and
NoiseReductionApplied tags, which may differ from the raw IFD versions of these tags due to sharpening or
noise reduction having been applied to the enhanced image data.
The enhanced IFD should have a non-null EnhanceParams tag to document the operations used to enhance
the image data.
Semantic Masks
DNG 1.6.0.0 and later supports the concept of "Semantic Masks" which may be used for purposes such as
image segmentation.
Semantic information is supported by allowing semantic masks (images) to be stored in IFDs with a
NewSubFileType of 65540 (10004.H). See section on NewSubFileType for details.
The semantic information IFD should use a PhotometricInterpretation of 52527 (CD2F.H). See section on
PhotometricInterpretation for details.
Semantic information is stored as a single channel image with one of the following pixel data types, with valid
encoding range indicated in parentheses:
• 8-bit unsigned integer (0 to 255)
• 16-bit unsigned integer (0 to 65535)
剩余122页未读，继续阅读
