usage.txt

This file was part of the Independent JPEG Group's software:
Copyright (C) 1991-2020, Thomas G. Lane, Guido Vollbeding.
libjpeg-turbo Modifications:
Copyright (C) 2010, 2012, 2014-2017, 2020-2024, D. R. Commander.
For conditions of distribution and use, see the accompanying README.ijg file.

USAGE instructions for the Independent JPEG Group's JPEG software
=================================================================

This file describes usage of the JPEG conversion programs cjpeg and djpeg,
as well as the utility programs jpegtran, rdjpgcom and wrjpgcom.  (See
the other documentation files if you wish to use the JPEG library within
your own programs.)

If you are on a Unix machine you may prefer to read the Unix-style manual
pages in files cjpeg.1, djpeg.1, jpegtran.1, rdjpgcom.1, wrjpgcom.1.


INTRODUCTION

These programs implement JPEG image encoding, decoding, and transcoding.
JPEG (pronounced "jay-peg") is a standardized compression method for
full-color and grayscale images.


GENERAL USAGE

We provide two programs, cjpeg to compress an image file into JPEG format,
and djpeg to decompress a JPEG file back into a conventional image format.

On most systems, you say:
        cjpeg [switches] [imagefile] >jpegfile
or
        djpeg [switches] [jpegfile]  >imagefile
The programs read the specified input file, or standard input if none is
named.  They always write to standard output (with trace/error messages to
standard error).  These conventions are handy for piping images between
programs.

If you defined TWO_FILE_COMMANDLINE when compiling the programs, you can
instead say:
        cjpeg [switches] imagefile jpegfile
or
        djpeg [switches] jpegfile  imagefile
i.e., both the input and output files are named on the command line.  This
style is a little more foolproof, and it loses no functionality if you don't
have pipes.

You can also say:
        cjpeg [switches] -outfile jpegfile  imagefile
or
        djpeg [switches] -outfile imagefile jpegfile
This syntax works on all systems, so it is useful for scripts.

The currently supported image file formats are: PPM (PBMPLUS color format),
PGM (PBMPLUS grayscale format), BMP, GIF [legacy feature], and Targa [legacy
feature].  cjpeg recognizes the input image format automatically, with the
exception of some Targa files.  You have to tell djpeg which format to
generate.

JPEG files are in the defacto standard JFIF file format.  There are other,
less widely used JPEG-based file formats, but we don't support them.

All switch names may be abbreviated; for example, -grayscale may be written
-gray or -gr.  Most of the "basic" switches can be abbreviated to as little as
one letter.  Upper and lower case are equivalent (-BMP is the same as -bmp).
British spellings are also accepted (e.g., -greyscale), though for brevity
these are not mentioned below.


CJPEG DETAILS

The basic command line switches for cjpeg are:

        -quality N[,...]  Scale quantization tables to adjust image quality.
                          Quality is 0 (worst) to 100 (best); default is 75.
                          (See below for more info.)

        -grayscale      Create monochrome JPEG file from color input.  By
                        specifying -grayscale, you'll get a smaller JPEG file
                        that takes less time to process.

        -rgb            Create RGB JPEG file.  Using this switch suppresses the
                        conversion from RGB colorspace input to the default
                        YCbCr JPEG colorspace.

        -optimize       Perform optimization of entropy encoding parameters.
                        Without this, default encoding parameters are used.
                        -optimize usually makes the JPEG file a little smaller,
                        but cjpeg runs somewhat slower and needs much more
                        memory.  Image quality and speed of decompression are
                        unaffected by -optimize.

        -progressive    Create progressive JPEG file (see below).  Implies
                        -optimize unless -arithmetic is also specified.

        -targa          Input file is Targa format [legacy feature].  Targa
                        files that contain an "identification" field will not
                        be automatically recognized by cjpeg.  For such files,
                        you must specify -targa to make cjpeg treat the input
                        as Targa format.  For most Targa files, you won't need
                        this switch.

The -quality switch lets you trade off compressed file size against quality of
the reconstructed image: the higher the quality setting, the larger the JPEG
file, and the closer the output image will be to the original input.  Normally
you want to use the lowest quality setting (smallest file) that decompresses
into something visually indistinguishable from the original image.  For this
purpose the quality setting should generally be between 50 and 95 (the default
is 75) for photographic images.  If you see defects at -quality 75, then go up
5 or 10 counts at a time until you are happy with the output image.  (The
optimal setting will vary from one image to another.)

-quality 100 will generate a quantization table of all 1's, minimizing loss
in the quantization step (but there is still information loss in subsampling,
as well as roundoff error.)  For most images, specifying a quality value above
about 95 will increase the size of the compressed file dramatically, and while
the quality gain from these higher quality values is measurable (using metrics
such as PSNR or SSIM), it is rarely perceivable by human vision.

In the other direction, quality values below 50 will produce very small files
of low image quality.  Settings around 5 to 10 might be useful in preparing an
index of a large image library, for example.  Try -quality 2 (or so) for some
amusing Cubist effects.  (Note: quality values below about 25 generate 2-byte
quantization tables, which are considered optional in the JPEG standard.
cjpeg emits a warning message when you give such a quality value, because some
other JPEG programs may be unable to decode the resulting file.  Use -baseline
if you need to ensure compatibility at low quality values.)

The -quality option has been extended in this version of cjpeg to support
separate quality settings for luminance and chrominance (or, in general,
separate settings for every quantization table slot.)  The principle is the
same as chrominance subsampling:  since the human eye is more sensitive to
spatial changes in brightness than spatial changes in color, the chrominance
components can be quantized more than the luminance components without
incurring any visible image quality loss.  However, unlike subsampling, this
feature reduces data in the frequency domain instead of the spatial domain,
which allows for more fine-grained control.  This option is useful in
quality-sensitive applications, for which the artifacts generated by
subsampling may be unacceptable.

The -quality option accepts a comma-separated list of parameters, which
respectively refer to the quality levels that should be assigned to the
quantization table slots.  If there are more q-table slots than parameters,
then the last parameter is replicated.  Thus, if only one quality parameter is
given, this is used for both luminance and chrominance (slots 0 and 1,
respectively), preserving the legacy behavior of cjpeg v6b and prior.  More (or
customized) quantization tables can be set with the -qtables option and
assigned to components with the -qslots option (see the "wizard" switches
below.)

JPEG files generated with separate luminance and chrominance quality are fully
compliant with standard JPEG decoders.

CAUTION: For this setting to be useful, be sure to pass an argument of
-sample 1x1 to cjpeg to disable chrominance subsampling.  Otherwise, the
default subsampling level (2x2, AKA "4:2:0") will be used.

The -progressive switch creates a "progressive JPEG" file.  In this type of
JPEG file, the data is stored in multiple scans of increasing quality.  If the
file is being transmitted over a slow communications link, the decoder can use
the first scan to display a low-quality image very quickly, and can then
improve the display with each subsequent scan.  The final image is exactly
equivalent to a standard JPEG file of the same quality setting, and the total
file size is about the same --- often a little smaller.

Switches for advanced users:

        -precision N   Create JPEG file with N-bit data precision.
                       N is 8, 12, or 16; default is 8.  If N is 16, then
                       -lossless must also be specified.  Note that only the
                       PBMPLUS input file format supports data precisions other
                       than 8.  (For historical reasons, cjpeg allows GIF input
                       files to be converted into 12-bit-per-sample JPEG files,
                       but this is not a useful conversion.)  Note also that
                       PBMPLUS input files are silently scaled to the target
                       data precision, even if it is lower than the precision
                       of the input file.  Passing an argument of -verbose to
                       cjpeg will cause it to print information about the
                       precision of the input file.  CAUTION: 12-bit and 16-bit
                       data precision is not yet widely implemented, so many
                       decoders will be unable to handle a 12-bit-per-sample or
                       16-bit-per-sample JPEG file at all.

                       "-precision 12" implies -optimize unless -arithmetic is
                       also specified.

        -lossless psv[,Pt] Create a lossless JPEG file using the specified
                        predictor selection value (1 - 7) and optional point
                        transform (0 - {precision}-1, where {precision} is the
                        JPEG data precision in bits).  A point transform value
                        of 0 (the default) is necessary in order to create a
                        fully lossless JPEG file.  (A non-zero point transform
                        value right-shifts the input samples by the specified
                        number of bits, which is effectively a form of lossy
                        color quantization.)  CAUTION: lossless JPEG is not yet
                        widely implemented, so many decoders will be unable to
                        handle a lossless JPEG file at all.  In most cases,
                        compressing and decompressing a lossless JPEG file is
                        considerably slower than compressing and decompressing
                        a lossy JPEG file, and lossless JPEG files are much
                        larger than lossy JPEG files.  Also note that the
                        following features will be unavailable when compressing
                        or decompressing a lossless JPEG file:
                          * Quality/quantization table selection
                          * Color space conversion (the JPEG image will use the
                            same color space as the input image)
                          * Color quantization
                          * DCT/IDCT algorithm selection
                          * Smoothing
                          * Downsampling/upsampling
                          * IDCT scaling
                          * Partial image decompression
                          * Transformations using jpegtran
                        Any switches used to enable or configure those features
                        will be ignored.

        -arithmetic     Use arithmetic coding.  CAUTION: arithmetic-coded JPEG
                        is not yet widely implemented, so many decoders will
                        be unable to handle an arithmetic-coded JPEG file at
                        all.

        -dct int        Use accurate integer DCT method (default).
        -dct fast       Use less accurate integer DCT method [legacy feature].
                        When the Independent JPEG Group's software was first
                        released in 1991, the compression time for a
                        1-megapixel JPEG image on a mainstream PC was measured
                        in minutes.  Thus, the fast integer DCT algorithm
                        provided noticeable performance benefits.  On modern
                        CPUs running libjpeg-turbo, however, the compression
                        time for a 1-megapixel JPEG image is measured in
                        milliseconds, and thus the performance benefits of the
                        fast algorithm are much less noticeable.  On modern
                        x86/x86-64 CPUs that support AVX2 instructions, the
                        fast and int methods have similar performance.  On
                        other types of CPUs, the fast method is generally about
                        5-15% faster than the int method.

                        For quality levels of 90 and below, there should be
                        little or no perceptible quality difference between the
                        two algorithms.  For quality levels above 90, however,
                        the difference between the fast and int methods becomes
                        more pronounced.  With quality=97, for instance, the
                        fast method incurs generally about a 1-3 dB loss in
                        PSNR relative to the int method, but this can be larger
                        for some images.  Do not use the fast method with
                        quality levels above 97.  The algorithm often
                        degenerates at quality=98 and above and can actually
                        produce a more lossy image than if lower quality levels
                        had been used.  Also, in libjpeg-turbo, the fast method
                        is not fully accelerated for quality levels above 97,
                        so it will be slower than the int method.
        -dct float      Use floating-point DCT method [legacy feature].
                        The float method does not produce significantly more
                        accurate results than the int method, and it is much
                        slower.  The float method may also give different
                        results on different machines due to varying roundoff
                        behavior, whereas the integer methods should give the
                        same results on all machines.

        -icc FILE       Embed ICC color management profile contained in the
                        specified file.

        -restart N      Emit a JPEG restart marker every N MCU rows, or every N
                        MCUs if "B" is attached to the number.

                        In typical JPEG images, an MCU (Minimum Coded Unit) is
                        the minimum set of interleaved "data units" (8x8 DCT
                        blocks if the image is lossy or samples if the image is
                        lossless) necessary to represent at least one data unit
                        per component.  (For example, an MCU in an interleaved
                        lossy JPEG image that uses 4:2:2 subsampling consists
                        of two luminance blocks followed by one block for each
                        chrominance component.)  In single-component or
                        non-interleaved JPEG images, an MCU is the same as a
                        data unit.  An MCU row is a row of MCUs spanning the
                        entire width of the image.

                        -restart 0 (the default) means no restart markers.

        -smooth N       Smooth the input image to eliminate dithering noise.
                        N, ranging from 1 to 100, indicates the strength of
                        smoothing.  0 (the default) means no smoothing.

        -maxmemory N    Set limit for amount of memory to use in processing
                        large images.  Value is in thousands of bytes, or
                        millions of bytes if "M" is attached to the number.
                        For example, -max 4m selects 4000000 bytes.  If more
                        space is needed, an error will occur.

        -memdst         Compress to memory instead of a file.  This feature was
                        implemented mainly as a way of testing the in-memory
                        destination manager (jpeg_mem_dest()), but it is also
                        useful for benchmarking, since it reduces the I/O
                        overhead.

        -report         Report compression progress.

        -strict         Treat all warnings as fatal.  Enabling this option will
                        cause the compressor to abort if an LZW-compressed GIF
                        input image contains incomplete or corrupt image data.

        -verbose        Enable debug printout.  More -v's give more output.
        or -debug       Also, version information is printed at startup.

        -version        Print version information and exit.

The -restart option inserts extra markers that allow a JPEG decoder to
resynchronize after a transmission error.  Without restart markers, any damage
to a compressed file will usually ruin the image from the point of the error
to the end of the image; with restart markers, the damage is usually confined
to the portion of the image up to the next restart marker.  Of course, the
restart markers occupy extra space.  We recommend -restart 1 for images that
will be transmitted across unreliable networks such as Usenet.

The -smooth option filters the input to eliminate fine-scale noise.  This is
often useful when converting dithered images to JPEG: a moderate smoothing
factor of 10 to 50 gets rid of dithering patterns in the input file, resulting
in a smaller JPEG file and a better-looking image.  Too large a smoothing
factor will visibly blur the image, however.

Switches for wizards:

        -baseline       Force baseline-compatible quantization tables to be
                        generated.  This clamps quantization values to 8 bits
                        even at low quality settings.  (This switch is poorly
                        named, since it does not ensure that the output is
                        actually baseline JPEG.  For example, you can use
                        -baseline and -progressive together.)

        -qtables file   Use the quantization tables given in the specified
                        text file.

        -qslots N[,...] Select which quantization table to use for each color
                        component.

        -sample HxV[,...]  Set JPEG sampling factors for each color component.

        -scans file     Use the scan script given in the specified text file.

The "wizard" switches are intended for experimentation with JPEG.  If you
don't know what you are doing, DON'T USE THEM.  These switches are documented
further in the file wizard.txt.


DJPEG DETAILS

The basic command line switches for djpeg are:

        -colors N       Reduce image to at most N colors [legacy feature].
        or -quantize N  This reduces the number of colors used in the output
                        image so that it can be stored in a colormapped file
                        format.  This feature cannot be used when decompressing
                        lossless JPEG images.  (-colors is the recommended
                        name.  -quantize is provided only for backward
                        compatibility.)

        -fast           Select recommended processing options for low-quality
                        output [legacy feature].  (The default options are
                        chosen for highest-quality output.)  Currently, this is
                        equivalent to "-dct fast -nosmooth -onepass -dither
                        ordered".  On modern CPUs, these settings have little
                        or no performance benefit and are retained solely for
                        backward compatibility.

        -grayscale      Force grayscale output even if JPEG file is full-color.
                        This feature cannot be used when decompressing
                        full-color lossless JPEG images.

        -rgb            Force RGB output even if JPEG file is grayscale.  This
                        feature cannot be used when decompressing grayscale
                        lossless JPEG images.

        -scale M/N      Scale the output image by a factor M/N.  Currently the
                        scale factor must be M/8, where M is an integer between
                        1 and 16 inclusive, or any reduced fraction thereof
                        (such as 1/2, 3/4, etc.)  Scaling is handy if the image
                        is larger than your screen.  This feature cannot be
                        used when decompressing lossless JPEG images.

        -bmp            Select BMP output format (Windows flavor).  8-bit
                        colormapped format is emitted if -colors or -grayscale
                        is specified, or if the JPEG file is grayscale;
                        otherwise, 24-bit full-color format is emitted.  This
                        format can only be used when decompressing
                        8-bit-per-sample JPEG images.

        -gif            Select GIF output format (LZW-compressed) [legacy
                        feature].  Since GIF does not support more than 256
                        colors, -colors 256 is assumed (unless you specify a
                        smaller number of colors).  If you specify -fast, the
                        default number of colors is 216.  This format can only
                        be used when decompressing 8-bit-per-sample or
                        12-bit-per-sample lossy JPEG images.

        -gif0           Select GIF output format (uncompressed) [legacy
                        feature].  Since GIF does not support more than 256
                        colors, -colors 256 is assumed (unless you specify a
                        smaller number of colors).  If you specify -fast, the
                        default number of colors is 216.  This format can only
                        be used when decompressing 8-bit-per-sample or
                        12-bit-per-sample lossy JPEG images.

        -os2            Select BMP output format (OS/2 1.x flavor) [legacy
                        feature].  8-bit colormapped format is emitted if
                        -colors or -grayscale is specified, or if the JPEG file
                        is grayscale; otherwise, 24-bit full-color format is
                        emitted.  This format can only be used when
                        decompressing 8-bit-per-sample JPEG images.

        -pnm            Select PBMPLUS (PPM/PGM) output format (this is the
                        default format).  PGM is emitted if the JPEG file is
                        grayscale or if -grayscale is specified; otherwise PPM
                        is emitted.

        -targa          Select Targa output format [legacy feature].  Grayscale
                        format is emitted if the JPEG file is grayscale or if
                        -grayscale is specified; otherwise, colormapped format
                        is emitted if -colors is specified; otherwise, 24-bit
                        full-color format is emitted.  This format can only be
                        used when decompressing 8-bit-per-sample JPEG images.

Switches for advanced users:

        -dct int        Use accurate integer DCT method (default).
        -dct fast       Use less accurate integer DCT method [legacy feature].
                        When the Independent JPEG Group's software was first
                        released in 1991, the decompression time for a
                        1-megapixel JPEG image on a mainstream PC was measured
                        in minutes.  Thus, the fast integer DCT algorithm
                        provided noticeable performance benefits.  On modern
                        CPUs running libjpeg-turbo, however, the decompression
                        time for a 1-megapixel JPEG image is measured in
                        milliseconds, and thus the performance benefits of the
                        fast algorithm are much less noticeable.  On modern
                        x86/x86-64 CPUs that support AVX2 instructions, the
                        fast and int methods have similar performance.  On
                        other types of CPUs, the fast method is generally about
                        5-15% faster than the int method.

                        If the JPEG image was compressed using a quality level
                        of 85 or below, then there should be little or no
                        perceptible quality difference between the two
                        algorithms.  When decompressing images that were
                        compressed using quality levels above 85, however, the
                        difference between the fast and int methods becomes
                        more pronounced.  With images compressed using
                        quality=97, for instance, the fast method incurs
                        generally about a 4-6 dB loss in PSNR relative to the
                        int method, but this can be larger for some images.  If
                        you can avoid it, do not use the fast method when
                        decompressing images that were compressed using quality
                        levels above 97.  The algorithm often degenerates for
                        such images and can actually produce a more lossy
                        output image than if the JPEG image had been compressed
                        using lower quality levels.
        -dct float      Use floating-point DCT method [legacy feature].
                        The float method does not produce significantly more
                        accurate results than the int method, and it is much
                        slower.  The float method may also give different
                        results on different machines due to varying roundoff
                        behavior, whereas the integer methods should give the
                        same results on all machines.

        -dither fs      Use Floyd-Steinberg dithering when quantizing colors
                        [legacy feature].
        -dither ordered Use ordered dithering when quantizing colors [legacy
                        feature].
        -dither none    Do not use dithering when quantizing colors [legacy
                        feature].  By default, Floyd-Steinberg dithering is
                        applied when quantizing colors.  This is slower but
                        usually produces the best results.  Ordered dithering
                        is a compromise between speed and quality.  No
                        dithering is faster but usually looks awful.  Note that
                        these switches have no effect unless color quantization
                        is being done.  Ordered dithering is only available in
                        -onepass mode.

        -icc FILE       Extract ICC color management profile to the specified
                        file.

        -map FILE       Quantize to the colors used in the specified image file
                        [legacy feature].  This is useful for producing
                        multiple files with identical color maps, or for
                        forcing a predefined set of colors to be used.  The
                        FILE must be a GIF or PPM file.  This option overrides
                        -colors and -onepass.

        -nosmooth       Use a faster, lower-quality upsampling routine.

        -onepass        Use one-pass instead of two-pass color quantization
                        [legacy feature].  The one-pass method needs less
                        memory, but it produces a lower-quality image.
                        -onepass is ignored unless you also specify -colors N.
                        Also, the one-pass method is always used for grayscale
                        output.  (The two-pass method has no improvement in
                        that case.)

        -maxmemory N    Set limit for amount of memory to use in processing
                        large images.  Value is in thousands of bytes, or
                        millions of bytes if "M" is attached to the number.
                        For example, -max 4m selects 4000000 bytes.  If more
                        space is needed, an error will occur.

        -maxscans N     Abort if the JPEG image contains more than N scans.
                        This feature demonstrates a method by which
                        applications can guard against denial-of-service
                        attacks instigated by specially-crafted malformed JPEG
                        images containing numerous scans with missing image
                        data or image data consisting only of "EOB runs" (a
                        feature of progressive JPEG images that allows
                        potentially hundreds of thousands of adjoining
                        zero-value pixels to be represented using only a few
                        bytes.)  Attempting to decompress such malformed JPEG
                        images can cause excessive CPU activity, since the
                        decompressor must fully process each scan (even if the
                        scan is corrupt) before it can proceed to the next
                        scan.

        -memsrc         Load input file into memory before decompressing.  This
                        feature was implemented mainly as a way of testing the
                        in-memory source manager (jpeg_mem_src().)

        -report         Report decompression progress.

        -skip Y0,Y1     Decompress all rows of the JPEG image except those
                        between Y0 and Y1 (inclusive.)  Note that if
                        decompression scaling is being used, then Y0 and Y1 are
                        relative to the scaled image dimensions.

        -crop WxH+X+Y   Decompress only a rectangular subregion of the image,
                        starting at point X,Y with width W and height H.  If
                        necessary, X will be shifted left to the nearest iMCU
                        boundary, and the width will be increased accordingly.
                        Note that if decompression scaling is being used, then
                        X, Y, W, and H are relative to the scaled image
                        dimensions.  Currently this option only works with the
                        PBMPLUS (PPM/PGM), GIF, and Targa output formats.

        -strict         Treat all warnings as fatal.  This feature also
                        demonstrates a method by which applications can guard
                        against attacks instigated by specially-crafted
                        malformed JPEG images.  Enabling this option will cause
                        the decompressor to abort if the JPEG image contains
                        incomplete or corrupt image data.

        -verbose        Enable debug printout.  More -v's give more output.
        or  -debug      Also, version information is printed at startup.

        -version        Print version information and exit.


HINTS FOR CJPEG

Color GIF files are not the ideal input for JPEG; JPEG is really intended for
compressing full-color (24-bit through 48-bit) images.  In particular, don't
try to convert cartoons, line drawings, and other images that have only a few
distinct colors.  GIF works great on these; JPEG does not.  If you want to
convert a GIF to JPEG, you should experiment with cjpeg's -quality and -smooth
options to get a satisfactory conversion.  -smooth 10 or so is often helpful.

Avoid running an image through a series of JPEG compression/decompression
cycles.  Image quality loss will accumulate; after ten or so cycles the image
may be noticeably worse than it was after one cycle.  It's best to use a
lossless format while manipulating an image, then convert to JPEG format when
you are ready to file the image away.

The -optimize option to cjpeg is worth using when you are making a "final"
version for posting or archiving.  It's also a win when you are using low
quality settings to make very small JPEG files; the percentage improvement
is often a lot more than it is on larger files.  (At present, -optimize
mode is always selected when generating progressive JPEG files.)


HINTS FOR BOTH PROGRAMS

If the memory needed by cjpeg or djpeg exceeds the limit specified by
-maxmemory, an error will occur.  You can leave out -progressive and -optimize
(for cjpeg) or specify -onepass (for djpeg) to reduce memory usage.

On machines that have "environment" variables, you can define the environment
variable JPEGMEM to set the default memory limit.  The value is specified as
described for the -maxmemory switch.  JPEGMEM overrides the default value
specified when the program was compiled, and itself is overridden by an
explicit -maxmemory switch.


JPEGTRAN

jpegtran performs various useful transformations of lossy (DCT-based) JPEG
files.  It can translate the coded representation from one variant of JPEG to
another, for example from baseline JPEG to progressive JPEG or vice versa.  It
can also perform some rearrangements of the image data, for example turning an
image from landscape to portrait format by rotation.  For EXIF files and JPEG
files containing Exif data, you may prefer to use exiftran instead.

jpegtran works by rearranging the compressed data (DCT coefficients), without
ever fully decoding the image.  Therefore, its transformations are lossless:
there is no image degradation at all, which would not be true if you used
djpeg followed by cjpeg to accomplish the same conversion.  But by the same
token, jpegtran cannot perform lossy operations such as changing the image
quality.  However, while the image data is losslessly transformed, metadata
can be removed.  See the -copy option for specifics.

jpegtran uses a command line syntax similar to cjpeg or djpeg.
On most systems, you say:
        jpegtran [switches] [inputfile] >outputfile
If you defined TWO_FILE_COMMANDLINE when compiling the program, you can instead
say:
        jpegtran [switches] inputfile outputfile
where both the input and output files are JPEG files.

To specify the coded JPEG representation used in the output file,
jpegtran accepts a subset of the switches recognized by cjpeg:
        -optimize       Perform optimization of entropy encoding parameters.
        -progressive    Create progressive JPEG file.
        -arithmetic     Use arithmetic coding.
        -restart N      Emit a JPEG restart marker every N MCU rows, or every
                        N MCUs if "B" is attached to the number.
        -scans file     Use the scan script given in the specified text file.
See the previous discussion of cjpeg for more details about these switches.
If you specify none of these switches, you get a plain baseline-JPEG output
file.  The quality setting and so forth are determined by the input file.

The image can be losslessly transformed by giving one of these switches:
        -flip horizontal        Mirror image horizontally (left-right).
        -flip vertical          Mirror image vertically (top-bottom).
        -rotate 90              Rotate image 90 degrees clockwise.
        -rotate 180             Rotate image 180 degrees.
        -rotate 270             Rotate image 270 degrees clockwise (or 90 ccw).
        -transpose              Transpose image (across UL-to-LR axis).
        -transverse             Transverse transpose (across UR-to-LL axis).

The transpose transformation has no restrictions regarding image dimensions.
The other transformations operate rather oddly if the image dimensions are not
a multiple of the iMCU size (usually 8 or 16 pixels), because they can only
transform complete blocks of DCT coefficient data in the desired way.

jpegtran's default behavior when transforming an odd-size image is designed
to preserve exact reversibility and mathematical consistency of the
transformation set.  As stated, transpose is able to flip the entire image
area.  Horizontal mirroring leaves any partial iMCU column at the right edge
untouched, but is able to flip all rows of the image.  Similarly, vertical
mirroring leaves any partial iMCU row at the bottom edge untouched, but is
able to flip all columns.  The other transforms can be built up as sequences
of transpose and flip operations; for consistency, their actions on edge
pixels are defined to be the same as the end result of the corresponding
transpose-and-flip sequence.

For practical use, you may prefer to discard any untransformable edge pixels
rather than having a strange-looking strip along the right and/or bottom edges
of a transformed image.  To do this, add the -trim switch:
        -trim           Drop non-transformable edge blocks.
Obviously, a transformation with -trim is not reversible, so strictly speaking
jpegtran with this switch is not lossless.  Also, the expected mathematical
equivalences between the transformations no longer hold.  For example,
"-rot 270 -trim" trims only the bottom edge, but "-rot 90 -trim" followed by
"-rot 180 -trim" trims both edges.

If you are only interested in perfect transformations, add the -perfect switch:
        -perfect        Fail with an error if the transformation is not
                        perfect.
For example, you may want to do
  jpegtran -rot 90 -perfect foo.jpg || djpeg foo.jpg | pnmflip -r90 | cjpeg
to do a perfect rotation, if available, or an approximated one if not.

This version of jpegtran also offers a lossless crop option, which discards
data outside of a given image region but losslessly preserves what is inside.
Like the rotate and flip transforms, lossless crop is restricted by the current
JPEG format; the upper left corner of the selected region must fall on an iMCU
boundary.  If it doesn't, then it is silently moved up and/or left to the
nearest iMCU boundary (the lower right corner is unchanged.)  Thus, the output
image covers at least the requested region, but it may cover more.  The
adjustment of the region dimensions may be optionally disabled by attaching an
'f' character ("force") to the width or height number.

The image can be losslessly cropped by giving the switch:
        -crop WxH+X+Y   Crop to a rectangular region of width W and height H,
                        starting at point X,Y.

If W or H is larger than the width/height of the input image, then the output
image is expanded in size, and the expanded region is filled in with zeros
(neutral gray).  Attaching an 'f' character ("flatten") to the width number
will cause each block in the expanded region to be filled in with the DC
coefficient of the nearest block in the input image rather than grayed out.
Attaching an 'r' character ("reflect") to the width number will cause the
expanded region to be filled in with repeated reflections of the input image
rather than grayed out.

A complementary lossless wipe option is provided to discard (gray out) data
inside a given image region while losslessly preserving what is outside:
        -wipe WxH+X+Y   Wipe (gray out) a rectangular region of width W and
                        height H from the input image, starting at point X,Y.

Attaching an 'f' character ("flatten") to the width number will cause the
region to be filled with the average of adjacent blocks rather than grayed out.
If the wipe region and the region outside the wipe region, when adjusted to the
nearest iMCU boundary, form two horizontally adjacent rectangles, then
attaching an 'r' character ("reflect") to the width number will cause the wipe
region to be filled with repeated reflections of the outside region rather than
grayed out.

A lossless drop option is also provided, which allows another JPEG image to be
inserted ("dropped") into the input image data at a given position, replacing
the existing image data at that position:
        -drop +X+Y filename     Drop (insert) another image at point X,Y

Both the input image and the drop image must have the same subsampling level.
It is best if they also have the same quantization (quality.)  Otherwise, the
quantization of the output image will be adapted to accommodate the higher of
the input image quality and the drop image quality.  The trim option can be
used with the drop option to requantize the drop image to match the input
image.  Note that a grayscale image can be dropped into a full-color image or
vice versa, as long as the full-color image has no vertical subsampling.  If
the input image is grayscale and the drop image is full-color, then the
chrominance channels from the drop image will be discarded.

Other not-strictly-lossless transformation switches are:

        -grayscale      Force grayscale output.
This option discards the chrominance channels if the input image is YCbCr
(ie, a standard color JPEG), resulting in a grayscale JPEG file.  The
luminance channel is preserved exactly, so this is a better method of reducing
to grayscale than decompression, conversion, and recompression.  This switch
is particularly handy for fixing a monochrome picture that was mistakenly
encoded as a color JPEG.  (In such a case, the space savings from getting rid
of the near-empty chroma channels won't be large; but the decoding time for
a grayscale JPEG is substantially less than that for a color JPEG.)

jpegtran also recognizes these switches that control what to do with "extra"
markers, such as comment blocks:
        -copy none      Copy no extra markers from source file.  This setting
                        suppresses all comments and other metadata in the
                        source file.
        -copy comments  Copy only comment markers.  This setting copies
                        comments from the source file but discards any other
                        metadata.
        -copy icc       Copy only ICC profile markers.  This setting copies the
                        ICC profile from the source file but discards any other
                        metadata.
        -copy all       Copy all extra markers.  This setting preserves
                        miscellaneous markers found in the source file, such
                        as JFIF thumbnails, Exif data, and Photoshop settings.
                        In some files, these extra markers can be sizable.
                        Note that this option will copy thumbnails as-is;
                        they will not be transformed.
The default behavior is -copy comments.  (Note: in IJG releases v6 and v6a,
jpegtran always did the equivalent of -copy none.)

Additional switches recognized by jpegtran are:
        -icc FILE
        -maxmemory N
        -maxscans N
        -outfile filename
        -report
        -strict
        -verbose
        -debug
        -version
These work the same as in cjpeg or djpeg.


THE COMMENT UTILITIES

The JPEG standard allows "comment" (COM) blocks to occur within a JPEG file.
Although the standard doesn't actually define what COM blocks are for, they
are widely used to hold user-supplied text strings.  This lets you add
annotations, titles, index terms, etc to your JPEG files, and later retrieve
them as text.  COM blocks do not interfere with the image stored in the JPEG
file.  The maximum size of a COM block is 64K, but you can have as many of
them as you like in one JPEG file.

We provide two utility programs to display COM block contents and add COM
blocks to a JPEG file.

rdjpgcom searches a JPEG file and prints the contents of any COM blocks on
standard output.  The command line syntax is
        rdjpgcom [-raw] [-verbose] [inputfilename]
The switch "-raw" (or just "-r") causes rdjpgcom to output non-printable
characters in JPEG comments.  These characters are normally escaped for
security reasons.
The switch "-verbose" (or just "-v") causes rdjpgcom to also display the JPEG
image dimensions.  If you omit the input file name from the command line,
the JPEG file is read from standard input.  (This may not work on some
operating systems, if binary data can't be read from stdin.)

wrjpgcom adds a COM block, containing text you provide, to a JPEG file.
Ordinarily, the COM block is added after any existing COM blocks, but you
can delete the old COM blocks if you wish.  wrjpgcom produces a new JPEG
file; it does not modify the input file.  DO NOT try to overwrite the input
file by directing wrjpgcom's output back into it; on most systems this will
just destroy your file.

The command line syntax for wrjpgcom is similar to cjpeg's.  On most systems,
it is
        wrjpgcom [switches] [inputfilename]
The output file is written to standard output.  The input file comes from
the named file, or from standard input if no input file is named.

If you defined TWO_FILE_COMMANDLINE when compiling the program, the syntax is:
        wrjpgcom [switches] inputfilename outputfilename
where both input and output file names must be given explicitly.

wrjpgcom understands three switches:
        -replace                 Delete any existing COM blocks from the file.
        -comment "Comment text"  Supply new COM text on command line.
        -cfile name              Read text for new COM block from named file.
(Switch names can be abbreviated.)  If you have only one line of comment text
to add, you can provide it on the command line with -comment.  The comment
text must be surrounded with quotes so that it is treated as a single
argument.  Longer comments can be read from a text file.

If you give neither -comment nor -cfile, then wrjpgcom will read the comment
text from standard input.  (In this case an input image file name MUST be
supplied, so that the source JPEG file comes from somewhere else.)  You can
enter multiple lines, up to 64KB worth.  Type an end-of-file indicator
(usually control-D or control-Z) to terminate the comment text entry.

wrjpgcom will not add a COM block if the provided comment string is empty.
Therefore -replace -comment "" can be used to delete all COM blocks from a
file.

These utility programs do not depend on the IJG JPEG library.  In
particular, the source code for rdjpgcom is intended as an illustration of
the minimum amount of code required to parse a JPEG file header correctly.