jpayne@68: .TH JPEGTRAN 1 "28 August 2019"
jpayne@68: .SH NAME
jpayne@68: jpegtran \- lossless transformation of JPEG files
jpayne@68: .SH SYNOPSIS
jpayne@68: .B jpegtran
jpayne@68: [
jpayne@68: .I options
jpayne@68: ]
jpayne@68: [
jpayne@68: .I filename
jpayne@68: ]
jpayne@68: .LP
jpayne@68: .SH DESCRIPTION
jpayne@68: .LP
jpayne@68: .B jpegtran
jpayne@68: performs various useful transformations of JPEG files.
jpayne@68: It can translate the coded representation from one variant of JPEG to another,
jpayne@68: for example from baseline JPEG to progressive JPEG or vice versa.  It can also
jpayne@68: perform some rearrangements of the image data, for example turning an image
jpayne@68: from landscape to portrait format by rotation.
jpayne@68: .PP
jpayne@68: For EXIF files and JPEG files containing Exif data, you may prefer to use
jpayne@68: .B exiftran
jpayne@68: instead.
jpayne@68: .PP
jpayne@68: .B jpegtran
jpayne@68: works by rearranging the compressed data (DCT coefficients), without
jpayne@68: ever fully decoding the image.  Therefore, its transformations are lossless:
jpayne@68: there is no image degradation at all, which would not be true if you used
jpayne@68: .B djpeg
jpayne@68: followed by
jpayne@68: .B cjpeg
jpayne@68: to accomplish the same conversion.  But by the same token,
jpayne@68: .B jpegtran
jpayne@68: cannot perform lossy operations such as changing the image quality.  However,
jpayne@68: while the image data is losslessly transformed, metadata can be removed.  See
jpayne@68: the
jpayne@68: .B \-copy
jpayne@68: option for specifics.
jpayne@68: .PP
jpayne@68: .B jpegtran
jpayne@68: reads the named JPEG/JFIF file, or the standard input if no file is
jpayne@68: named, and produces a JPEG/JFIF file on the standard output.
jpayne@68: .SH OPTIONS
jpayne@68: All switch names may be abbreviated; for example,
jpayne@68: .B \-optimize
jpayne@68: may be written
jpayne@68: .B \-opt
jpayne@68: or
jpayne@68: .BR \-o .
jpayne@68: Upper and lower case are equivalent.
jpayne@68: British spellings are also accepted (e.g.,
jpayne@68: .BR \-optimise ),
jpayne@68: though for brevity these are not mentioned below.
jpayne@68: .PP
jpayne@68: To specify the coded JPEG representation used in the output file,
jpayne@68: .B jpegtran
jpayne@68: accepts a subset of the switches recognized by
jpayne@68: .BR cjpeg :
jpayne@68: .TP
jpayne@68: .B \-optimize
jpayne@68: Perform optimization of entropy encoding parameters.
jpayne@68: .TP
jpayne@68: .B \-progressive
jpayne@68: Create progressive JPEG file.
jpayne@68: .TP
jpayne@68: .BI \-restart " N"
jpayne@68: Emit a JPEG restart marker every N MCU rows, or every N MCU blocks if "B" is
jpayne@68: attached to the number.
jpayne@68: .TP
jpayne@68: .B \-arithmetic
jpayne@68: Use arithmetic coding.
jpayne@68: .TP
jpayne@68: .BI \-scans " file"
jpayne@68: Use the scan script given in the specified text file.
jpayne@68: .PP
jpayne@68: See
jpayne@68: .BR cjpeg (1)
jpayne@68: for more details about these switches.
jpayne@68: If you specify none of these switches, you get a plain baseline-JPEG output
jpayne@68: file.  The quality setting and so forth are determined by the input file.
jpayne@68: .PP
jpayne@68: The image can be losslessly transformed by giving one of these switches:
jpayne@68: .TP
jpayne@68: .B \-flip horizontal
jpayne@68: Mirror image horizontally (left-right).
jpayne@68: .TP
jpayne@68: .B \-flip vertical
jpayne@68: Mirror image vertically (top-bottom).
jpayne@68: .TP
jpayne@68: .B \-rotate 90
jpayne@68: Rotate image 90 degrees clockwise.
jpayne@68: .TP
jpayne@68: .B \-rotate 180
jpayne@68: Rotate image 180 degrees.
jpayne@68: .TP
jpayne@68: .B \-rotate 270
jpayne@68: Rotate image 270 degrees clockwise (or 90 ccw).
jpayne@68: .TP
jpayne@68: .B \-transpose
jpayne@68: Transpose image (across UL-to-LR axis).
jpayne@68: .TP
jpayne@68: .B \-transverse
jpayne@68: Transverse transpose (across UR-to-LL axis).
jpayne@68: .IP
jpayne@68: The transpose transformation has no restrictions regarding image dimensions.
jpayne@68: The other transformations operate rather oddly if the image dimensions are not
jpayne@68: a multiple of the iMCU size (usually 8 or 16 pixels), because they can only
jpayne@68: transform complete blocks of DCT coefficient data in the desired way.
jpayne@68: .IP
jpayne@68: .BR jpegtran 's
jpayne@68: default behavior when transforming an odd-size image is designed
jpayne@68: to preserve exact reversibility and mathematical consistency of the
jpayne@68: transformation set.  As stated, transpose is able to flip the entire image
jpayne@68: area.  Horizontal mirroring leaves any partial iMCU column at the right edge
jpayne@68: untouched, but is able to flip all rows of the image.  Similarly, vertical
jpayne@68: mirroring leaves any partial iMCU row at the bottom edge untouched, but is
jpayne@68: able to flip all columns.  The other transforms can be built up as sequences
jpayne@68: of transpose and flip operations; for consistency, their actions on edge
jpayne@68: pixels are defined to be the same as the end result of the corresponding
jpayne@68: transpose-and-flip sequence.
jpayne@68: .IP
jpayne@68: For practical use, you may prefer to discard any untransformable edge pixels
jpayne@68: rather than having a strange-looking strip along the right and/or bottom edges
jpayne@68: of a transformed image.  To do this, add the
jpayne@68: .B \-trim
jpayne@68: switch:
jpayne@68: .TP
jpayne@68: .B \-trim
jpayne@68: Drop non-transformable edge blocks.
jpayne@68: .IP
jpayne@68: Obviously, a transformation with
jpayne@68: .B \-trim
jpayne@68: is not reversible, so strictly speaking
jpayne@68: .B jpegtran
jpayne@68: with this switch is not lossless.  Also, the expected mathematical
jpayne@68: equivalences between the transformations no longer hold.  For example,
jpayne@68: .B \-rot 270 -trim
jpayne@68: trims only the bottom edge, but
jpayne@68: .B \-rot 90 -trim
jpayne@68: followed by
jpayne@68: .B \-rot 180 -trim
jpayne@68: trims both edges.
jpayne@68: .IP
jpayne@68: If you are only interested in perfect transformation, add the
jpayne@68: .B \-perfect
jpayne@68: switch:
jpayne@68: .TP
jpayne@68: .B \-perfect
jpayne@68: Fails with an error if the transformation is not perfect.
jpayne@68: .IP
jpayne@68: For example you may want to do
jpayne@68: .IP
jpayne@68: .B (jpegtran \-rot 90 -perfect
jpayne@68: .I foo.jpg
jpayne@68: .B || djpeg
jpayne@68: .I foo.jpg
jpayne@68: .B | pnmflip \-r90 | cjpeg)
jpayne@68: .IP
jpayne@68: to do a perfect rotation if available or an approximated one if not.
jpayne@68: .PP
jpayne@68: We also offer a lossless-crop option, which discards data outside a given
jpayne@68: image region but losslessly preserves what is inside.  Like the rotate and
jpayne@68: flip transforms, lossless crop is restricted by the current JPEG format: the
jpayne@68: upper left corner of the selected region must fall on an iMCU boundary.  If
jpayne@68: this does not hold for the given crop parameters, we silently move the upper
jpayne@68: left corner up and/or left to make it so, simultaneously increasing the
jpayne@68: region dimensions to keep the lower right crop corner unchanged.  (Thus, the
jpayne@68: output image covers at least the requested region, but may cover more.)
jpayne@68: The adjustment of the region dimensions may be optionally disabled by
jpayne@68: attaching an 'f' character ("force") to the width or height number.
jpayne@68: .PP
jpayne@68: The image can be losslessly cropped by giving the switch:
jpayne@68: .TP
jpayne@68: .B \-crop WxH+X+Y
jpayne@68: Crop to a rectangular subarea of width W, height H starting at point X,Y.
jpayne@68: .PP
jpayne@68: Crop extension: The width or height parameters can be made larger than the
jpayne@68: source image.  In this case the extra area is filled in with zero (neutral
jpayne@68: gray).  A larger width parameter has two more options:  Attaching an 'f'
jpayne@68: character ("flatten") to the width number will fill in the extra area with
jpayne@68: the DC of the adjacent block, instead of gray out.  Attaching an 'r'
jpayne@68: character ("reflect") to the width number will fill in the extra area with
jpayne@68: repeated reflections of the source region, instead of gray out.
jpayne@68: .PP
jpayne@68: A complementary lossless-wipe option is provided to discard (gray out) data
jpayne@68: inside a given image region while losslessly preserving what is outside:
jpayne@68: .TP
jpayne@68: .B \-wipe WxH+X+Y
jpayne@68: Wipe (gray out) a rectangular subarea of width W, height H starting at point
jpayne@68: X,Y.
jpayne@68: .PP
jpayne@68: Attaching an 'f' character ("flatten") to the width number will fill the
jpayne@68: region with the average of adjacent blocks, instead of gray out.  In case
jpayne@68: the wipe region and outside area form two horizontally adjacent rectangles,
jpayne@68: attaching an 'r' character ("reflect") to the width number will fill the
jpayne@68: region with repeated reflections of the outside area, instead of gray out.
jpayne@68: .PP
jpayne@68: Another option is lossless-drop, which replaces data at a given image
jpayne@68: position by another image:
jpayne@68: .TP
jpayne@68: .B \-drop +X+Y filename
jpayne@68: Drop another image
jpayne@68: .PP
jpayne@68: Both source images must have the same subsampling values.  It is best if
jpayne@68: they also have the same quantization, otherwise quantization adaption occurs.
jpayne@68: The trim option can be used with the drop option to requantize the drop file
jpayne@68: to the source file.
jpayne@68: .PP
jpayne@68: Other not-strictly-lossless transformation switches are:
jpayne@68: .TP
jpayne@68: .B \-grayscale
jpayne@68: Force grayscale output.
jpayne@68: .IP
jpayne@68: This option discards the chrominance channels if the input image is YCbCr
jpayne@68: (ie, a standard color JPEG), resulting in a grayscale JPEG file.  The
jpayne@68: luminance channel is preserved exactly, so this is a better method of reducing
jpayne@68: to grayscale than decompression, conversion, and recompression.  This switch
jpayne@68: is particularly handy for fixing a monochrome picture that was mistakenly
jpayne@68: encoded as a color JPEG.  (In such a case, the space savings from getting rid
jpayne@68: of the near-empty chroma channels won't be large; but the decoding time for
jpayne@68: a grayscale JPEG is substantially less than that for a color JPEG.)
jpayne@68: .TP
jpayne@68: .BI \-scale " M/N"
jpayne@68: Scale the output image by a factor M/N.
jpayne@68: .IP
jpayne@68: Currently supported scale factors are M/N with all M from 1 to 16, where N is
jpayne@68: the source DCT size, which is 8 for baseline JPEG.  If the /N part is omitted,
jpayne@68: then M specifies the DCT scaled size to be applied on the given input.  For
jpayne@68: baseline JPEG this is equivalent to M/8 scaling, since the source DCT size
jpayne@68: for baseline JPEG is 8.
jpayne@68: .B Caution:
jpayne@68: An implementation of the JPEG SmartScale extension is required for this
jpayne@68: feature.  SmartScale enabled JPEG is not yet widely implemented, so many
jpayne@68: decoders will be unable to view a SmartScale extended JPEG file at all.
jpayne@68: .PP
jpayne@68: .B jpegtran
jpayne@68: also recognizes these switches that control what to do with "extra" markers,
jpayne@68: such as comment blocks:
jpayne@68: .TP
jpayne@68: .B \-copy none
jpayne@68: Copy no extra markers from source file.  This setting suppresses all
jpayne@68: comments and other metadata in the source file.
jpayne@68: .TP
jpayne@68: .B \-copy comments
jpayne@68: Copy only comment markers.  This setting copies comments from the source file,
jpayne@68: but discards any other metadata.
jpayne@68: .TP
jpayne@68: .B \-copy all
jpayne@68: Copy all extra markers.  This setting preserves metadata
jpayne@68: found in the source file, such as JFIF thumbnails, Exif data, and Photoshop
jpayne@68: settings.  In some files these extra markers can be sizable.  Note that this
jpayne@68: option will copy thumbnails as-is; they will not be transformed.
jpayne@68: .IP
jpayne@68: The default behavior is
jpayne@68: .BR "\-copy comments" .
jpayne@68: (Note: in IJG releases v6 and v6a,
jpayne@68: .B jpegtran
jpayne@68: always did the equivalent of
jpayne@68: .BR "\-copy none" .)
jpayne@68: .PP
jpayne@68: Additional switches recognized by jpegtran are:
jpayne@68: .TP
jpayne@68: .BI \-maxmemory " N"
jpayne@68: Set limit for amount of memory to use in processing large images.  Value is
jpayne@68: in thousands of bytes, or millions of bytes if "M" is attached to the
jpayne@68: number.  For example,
jpayne@68: .B \-max 4m
jpayne@68: selects 4000000 bytes.  If more space is needed, temporary files will be used.
jpayne@68: .TP
jpayne@68: .BI \-outfile " name"
jpayne@68: Send output image to the named file, not to standard output.
jpayne@68: .TP
jpayne@68: .B \-verbose
jpayne@68: Enable debug printout.  More
jpayne@68: .BR \-v 's
jpayne@68: give more output.  Also, version information is printed at startup.
jpayne@68: .TP
jpayne@68: .B \-debug
jpayne@68: Same as
jpayne@68: .BR \-verbose .
jpayne@68: .SH EXAMPLES
jpayne@68: .LP
jpayne@68: This example converts a baseline JPEG file to progressive form:
jpayne@68: .IP
jpayne@68: .B jpegtran \-progressive
jpayne@68: .I foo.jpg
jpayne@68: .B >
jpayne@68: .I fooprog.jpg
jpayne@68: .PP
jpayne@68: This example rotates an image 90 degrees clockwise, discarding any
jpayne@68: unrotatable edge pixels:
jpayne@68: .IP
jpayne@68: .B jpegtran \-rot 90 -trim
jpayne@68: .I foo.jpg
jpayne@68: .B >
jpayne@68: .I foo90.jpg
jpayne@68: .SH ENVIRONMENT
jpayne@68: .TP
jpayne@68: .B JPEGMEM
jpayne@68: If this environment variable is set, its value is the default memory limit.
jpayne@68: The value is specified as described for the
jpayne@68: .B \-maxmemory
jpayne@68: switch.
jpayne@68: .B JPEGMEM
jpayne@68: overrides the default value specified when the program was compiled, and
jpayne@68: itself is overridden by an explicit
jpayne@68: .BR \-maxmemory .
jpayne@68: .SH SEE ALSO
jpayne@68: .BR cjpeg (1),
jpayne@68: .BR djpeg (1),
jpayne@68: .BR rdjpgcom (1),
jpayne@68: .BR wrjpgcom (1)
jpayne@68: .br
jpayne@68: Wallace, Gregory K.  "The JPEG Still Picture Compression Standard",
jpayne@68: Communications of the ACM, April 1991 (vol. 34, no. 4), pp. 30-44.
jpayne@68: .SH AUTHOR
jpayne@68: Independent JPEG Group
jpayne@68: .SH BUGS
jpayne@68: The transform options can't transform odd-size images perfectly.  Use
jpayne@68: .B \-trim
jpayne@68: or
jpayne@68: .B \-perfect
jpayne@68: if you don't like the results.
jpayne@68: .PP
jpayne@68: The entire image is read into memory and then written out again, even in
jpayne@68: cases where this isn't really necessary.  Expect swapping on large images,
jpayne@68: especially when using the more complex transform options.