image%02d.miff

    redslope,redoffset,redpower:
    greenslope,greenoffset,greenpower:
    blueslope,blueoffset,bluepower:
    saturation

    "1.0,0.0,1.0:1.0,0.0,1.0:1.0,0.0,1.0:1.0"

    gm convert -clip -negate cockatoo.tif negated.tif

    gm convert input.ppm -modulate 115,0,100 \
              -colorize 7,21,50 output.ppm.

     -comment "%m:%f %wx%h"

     Over
     In
     Out
     Atop
     Xor
     Plus
     Minus
     Add
     Subtract
     Difference
     Divide
     Multiply
     Bumpmap
     Copy
     CopyRed
     CopyGreen
     CopyBlue
     CopyOpacity
     CopyCyan
     CopyMagenta
     CopyYellow
     CopyBlack

Over

The result will be the union of the two image shapes, with opaque areas of change-image obscuring base-image in the region of overlap.

In

The result is simply change-image cut by the shape of base-image. None of the image data of base-image will be in the result.

Out

The resulting image is change-image with the shape of base-image cut out.

Atop

The result is the same shape as base-image, with change-image obscuring base-image where the image shapes overlap. Note this differs from over because the portion of change-image outside base-image's shape does not appear in the result.

Xor

The result is the image data from both change-image and base-image that is outside the overlap region. The overlap region will be blank.

Plus

The result is just the sum of the image data. Output values are cropped to MaxRGB (no overflow). This operation is independent of the matte channels.

Minus

The result of change-image - base-image, with underflow cropped to zero. The matte channel is ignored (set to opaque, full coverage).

Add

The result of change-image + base-image, with overflow wrapping around (mod MaxRGB+1).

Subtract

The result of change-image - base-image, with underflow wrapping around (mod MaxRGB+1). The add and subtract operators can be used to perform reversible transformations.

Difference

The result of abs(change-image - base-image). This is useful for comparing two very similar images.

Divide

The result of change-image / base-image. This is useful for improving the readability of text on unevenly illuminated photos (by using a gaussian blurred copy of change-image as base-image).

Multiply

The result of change-image * base-image. This is useful for the creation of drop-shadows.

Bumpmap

The result base-image shaded by change-image.

Copy

The resulting image is base-image replaced with change-image. Here the matte information is ignored.

CopyRed

The resulting image is the red channel in base-image replaced with the red channel in change-image. The other channels are copied untouched.

CopyGreen

The resulting image is the green channel in base-image replaced with the green channel in change-image. The other channels are copied untouched.

CopyBlue

The resulting image is the blue channel in base-image replaced with the blue channel in change-image. The other channels are copied untouched.

CopyOpacity

The resulting image is the opacity channel in base-image replaced with the opacity channel in change-image. The other channels are copied untouched.

CopyCyan

The resulting image is the cyan channel in base-image replaced with the cyan channel in change-image. The other channels are copied untouched. Use of this operator requires that base-image be in CMYK(A) colorspace.

CopyMagenta

The resulting image is the magenta channel in base-image replaced with the magenta channel in change-image. The other channels are copied untouched. Use of this operator requires that base-image be in CMYK(A) colorspace.

CopyYellow

The resulting image is the yellow channel in base-image replaced with the yellow channel in change-image. The other channels are copied untouched. Use of this operator requires that base-image be in CMYK(A) colorspace.

CopyBlack

The resulting image is the black channel in base-image replaced with the black channel in change-image. The other channels are copied untouched. Use of this operator requires that base-image be in CMYK(A) colorspace. If change-image is not in CMYK space, then the change-image pixel intensities are used.

    gm convert rose: -contrast -contrast rose_c2.png

    gm convert -debug "Cache,Blob" rose: rose.png

bmp:allow-jpeg={true|false}

If the bmp:allow-jpeg value is set to true, then enable BMP files using JPEG compression, under control of the compression option (which may be JPEG already if the input file is JPEG). JPEG-compressed BMP files are very rare so this option acts as a safeguard to assure that the current compression setting does not accidentally produce JPEG-compressed BMP files which most other software is unable to read.

cineon:colorspace={rgb|cineonlog}

Use the cineon:colorspace option when reading a Cineon file to specify the colorspace the Cineon file uses. This overrides the colorspace type implied by the DPX header (if any).

dpx:bits-per-sample=<value>

If the dpx:bits-per-sample key is defined, GraphicsMagick will write DPX images with the specified bits per sample, overriding any existing depth value. If this option is not specified, then the value is based on the existing image depth value from the original image file. The DPX standard supports bits per sample values of 1, 8, 10, 12, and 16. Many DPX readers demand a sample size of 10 bits with type A padding (see below).

dpx:colorspace={rgb|cineonlog}

Use the dpx:colorspace option when reading a DPX file to specify the colorspace the DPX file uses. This overrides the colorspace type implied by the DPX header (if any).

dpx:packing-method={packed|a|b|lsbpad|msbpad}

DPX samples are output within 32-bit words. They may be tightly packed end-to-end within the words ("packed"), padded with null bits to the right of the sample ("a" or "lsbpad"), or padded with null bits to the left of the sample ("b" or "msbpad"). This option only has an effect for sample sizes of 10 or 12 bits. If samples are not packed, the DPX standard recommends type A padding. Many DPX readers demand a sample size of 10 bits with type A padding.

dpx:pixel-endian={lsb|msb}

Allows the user to specify the endian order of the pixels when reading or writing the DPX files. Sometimes this is useful if the file is (or must be) written incorrectly so that the file header and the pixels use different endianness.

dpx:swap-samples={true|false}

dpx:swap-samples-read={true|false}

dpx:swap-samples-write={true|false}

GraphicsMagick strives to adhere to the DPX standard but certain aspects of the standard can be quite confusing. As a result, some 10-bit DPX files have Red and Blue interchanged, or Cb and Cr interchanged due to an different interpretation of the standard, or getting the wires crossed. The swap-samples option may be supplied when reading or writing in order to read or write using the necessary sample order. Use swap-samples-read when swapping should only occur in the reader, or swap-samples-write when swapping should only occur in the writer.

heif:ignore-transformations={true|false}

Return HEIF (e.g. HEIC/AVIF) image without any transformations (e.g. rotation) applied.

gradient:direction={South|North|West|East|NorthWest|NorthEast|SouthWest|SouthEast}

By default, the gradient coder produces a gradient from top to bottom ("South"). Since GraphicsMagick 1.3.35, the gradient direction may be specified to produce gradient vectors according to a gravity-like specification. The arguments are South (Top to Bottom), North (Bottom to Top), West (Right to Left), East (Left to Right), NorthWest (Bottom-Right to Top-Left), NorthEast (Bottom-Left to Top-Right), SouthWest (Top-Right Bottom-Left), and SouthEast (Top-Left to Bottom-Right).

jp2:rate=<value>

Specify the compression factor to use while writing JPEG-2000 files. The compression factor is the reciprocal of the compression ratio. The valid range is 0.0 to 1.0, with 1.0 indicating lossless compression. If defined, this value overrides the -quality setting. The default quality setting of 75 results in a rate value of 0.06641.

jpeg:arithmetic-coding={true|false}

Enables or disables arithmetic encoding if the JPEG library supports it (default disabled). When this is enabled, optimize-coding is necessarily disabled.

jpeg:block-smoothing={true|false}

Enables or disables block smoothing when reading a JPEG file (default enabled).

jpeg:dct-method=<value>

Selects the IJG JPEG library DCT implementation to use. The encoding implementations vary in speed and encoding error. The available choices for value are islow, ifast, float, default and fastest. Note that fastest might not necessarily be fastest on your CPU, depending on the choices made when the JPEG library was built and how your CPU behaves.

jpeg:fancy-upsampling={true|false}

Enables or disables fancy upsampling when reading a JPEG file (default enabled).

jpeg:data-precision=<value>

Selects the JPEG sample storage size. Normal JPEG uses 8. Turbo-jpeg and some variants of IJG JPEG also support 12 and 16 bits.

jpeg:lossless-precision=<value>

Selects a degree of precision to use while performing DPCM-based compression. The available range 0 to precision-1. For example, if the output depth is 16, then the available range is 0 to 15. When the precision is non-zero, then the pixels are right-shifted to effectively support fewer bits (fewer colors), but offering an improved compression ratio. Use a value of 0 to achieve lossless compression.

jpeg:lossless-predictor=<value>

Selects the predictor to use for lossless compression. The supported values are in the range of 1 to 7 and select a pre-defined predictor algorithm to use, which considers surrounding pixels. The best predictor for a particular image depends on the image.

jpeg:max-scan-number=<value>

Specifies an integer value for the maximum number of progressive scans allowed in a JPEG file. The default maximum is 100 scans. This limit is imposed due to a weakness in the JPEG standard which allows small JPEG files to take many minutes or hours to be read.

jpeg:max-warnings=<value>

Specifies an integer value for how many warnings are allowed for any given error type before being promoted to a hard error. JPEG files producing excessive warnings indicate a problem with the file.

jpeg:optimize-coding={true|false}

Selects if huffman encoding should be used. Huffman encoding is enabled by default, but may be disabled for very large images since it encoding requires that the entire image be buffered in memory. Huffman encoding produces smaller JPEG files at the expense of added compression time and memory consumption.

jpeg:preserve-settings

If the jpeg:preserve-settings flag is defined, the JPEG encoder will use the same "quality" and "sampling-factor" settings that were found in the input file, if the input was in JPEG format. These settings are also preserved if the input is a JPEG file and the output is a JNG file. If the colorspace of the output file differs from that of the input file, the quality setting is preserved but the sampling-factors are not.

pcl:fit-to-page

If the pcl:fit-to-page flag is defined, then the printer is requested to scale the image to fit the page size (width and/or height).

png:chunk-malloc-max=<value>

png:chunk-malloc-max specifies the maximum chunk size that libpng will be allowed to read. Libpng's default is normally 8,000,000 bytes. Very rarely, a valid PNG file may be encountered where the error is reported "chunk data is too large". In this case, the limit may be increased using this option. Take care when increasing this limit since an excessively large limit could allow untrusted files to use excessive memory.

mng:maximum-loops=<value>

mng:maximum-loops specifies the maximum number of loops allowed to be specified by a MNG LOOP chunk. Without an imposed limit, a MNG file could request up to 2147483647 loops, which could run for a very long time. The current default limit is 512 loops.

pdf:use-cropbox={true|false}

If the pdf:use-cropbox flag is set to true, then Ghostscript is requested to apply the PDF crop box.

pdf:stop-on-error={true|false}

If the pdf:stop-on-error flag is set to true, then Ghostscript is requested to stop processing the PDF when the first error is encountered. Otherwise it will attempt to process all requested pages.

ps:imagemask

If the ps:imagemask flag is defined, the PS3 and EPS3 coders will create Postscript files that render bilevel images with the Postscript imagemask operator instead of the image operator.

ptif:minimum-geometry=<geometry>

If the ptif:minimum-geometry key is defined, GraphicsMagick will use it to determine the minimum frame size to output when writing a pyramid TIFF file (a TIFF file containing a succession of reduced versions of the first frame). The default minimum geometry is 32x32.

tiff:alpha={unspecified|associated|unassociated}

Specify the TIFF alpha channel type when reading or writing TIFF files, overriding the normal value. The default alpha channel type for new files is unspecified alpha. Existing alpha settings are preserved when converting from one TIFF file to another. When a TIFF file uses associated alpha, the image pixels are pre-multiplied (i.e. altered) with the alpha channel. Files with "associated" alpha appear as if they were alpha composited on a black background when the matte channel is disabled. If the unassociated alpha type is selected, then the alpha channel is saved without altering the pixels. Photoshop recognizes associated alpha as transparency information, if the file is saved with unassociated alpha, the alpha information is loaded as an independent channel. Note that for many years, ImageMagick and GraphicsMagick marked TIFF files as using associated alpha, without properly pre-multiplying the pixels.

tiff:fill-order={msb2lsb|lsb2msb}

If the tiff:fill-order key is defined, GraphicsMagick will use it to determine the bit fill order used while writing TIFF files. The normal default is "msb2lsb", which matches the native bit order of all modern CPUs. The only exception to this is when Group3 or Group4 FAX compression is requested since FAX machines send data in bit-reversed order and therefore RFC 2301 recommends using reverse order.

tiff:group-three-options=<value>

If the tiff:group-three-options key is defined, GraphicsMagick will use it to set the group3 options tag when writing group3-compressed TIFF. Please see the TIFF specification for the usage of this tag. The default value is 4.

tiff:ignore-tags=<tags>

If the tiff:ignore-tags key is defined, then it is used as a list of comma-delimited integer TIFF tag values to ignore while reading the TIFF file. This is useful in order to be able to read files which which otherwise fail to read due to problems with TIFF tags. Note that some TIFF tags are required in order to be able to read the image data at all.

tiff:photometric={minisblack|miniswhite}

For grayscale and bilevel images, GraphicsMagick normally chooses the photometric based on bit depth and compression. If the tiff:photometric key is defined, GraphicsMagick will use it to determine the photometric used while writing grayscale and bilevel images. Normally bilevel images are miniswhite and grayscale images are minisblack.

tiff:report-warnings={false|true}

If the tiff:report-warnings key is defined and set to true, then TIFF warnings are reported as a warning exception rather than as a coder log message. Such warnings are reported after the image has been read or written. Most TIFF warnings are benign but sometimes they may help deduce problems with the TIFF file, or help detect that the TIFF file requires a special application to read successfully due to the use of proprietary or specialized extensions.

tiff:sample-format={unsigned|ieeefp}

If the tiff:sample-format key is defined, GraphicsMagick will use it to determine the sample format used while writing TIFF files. The default is "unsigned". Specify "ieeefp" in order to write floating-point TIFF files with float (32-bit) or double (64-bit) values. Use the tiff:bits-per-sample define to determine the type of floating-point value to use.

tiff:max-sample-value=<value>

If the tiff:max-sample-value key is defined, GraphicsMagick will use the assigned value as the maximum floating point value while reading or writing IEEE floating point TIFFs. Otherwise the maximum value is 1.0 or the value obtained from the file's SMaxSampleValue tag (if present). The floating point data is currently not scanned in advance to determine a best maximum sample value so if the range is not 1.0, or the SMaxSampleValue tag is not present, it may be necessary to (intelligently) use this parameter to properly read a file.

tiff:min-sample-value=<value>

If the tiff:min-sample-value key is defined, GraphicsMagick will use the assigned value as the minimum floating point value while reading or writing IEEE floating point TIFFs. Otherwise the minimum value is 0.0 or the value obtained from the file's SMinSampleValue tag (if present).

tiff:bits-per-sample=<value>

If the tiff:bits-per-sample key is defined, GraphicsMagick will write images with the specified bits per sample, overriding any existing depth value. Value may be any in the range of 1 to 32, or 64 when the default 'unsigned' format is written, or 16/32/24/64 if IEEEFP format is written. Please note that the baseline TIFF 6.0 specification only requires readers to handle certain powers of two, and the values to be handled depend on the nature of the image (e.g. colormapped, grayscale, RGB, CMYK).

tiff:samples-per-pixel=<value>

If the tiff:samples-per-pixel key is defined to a value, the TIFF coder will write TIFF images with the defined samples per pixel, overriding any value stored in the image. This option should not normally be used.

tiff:rows-per-strip=<value>

Allows the user to specify the number of rows per TIFF strip. Rounded up to a multiple of 16 when using JPEG compression. Ignored when using tiles.

tiff:strip-per-page=true

Requests that the image is written in a single TIFF strip. This is normally the default when group3 or group4 compression is requested within reasonable limits. Requesting a single strip for large images may result in failure due to resource consumption in the writer or reader.

tiff:tile

Enable writing tiled TIFF (rather than stripped) using the default tile size. Tiled TIFF organizes the image as an array of smaller images (tiles) in order to enable random access.

tiff:tile-geometry=<width>x<height>

Specify the tile size to use while writing tiled TIFF. Width and height should be a multiple of 16. If the value is not a multiple of 16, then it will be rounded down. Enables tiled TIFF if it has not already been enabled. GraphicsMagick does not use tiled storage internally so tiles need to be converted back and forth from the internal scanline-oriented storage to tile-oriented storage. Testing with typical RGB images shows that useful square tile size values range from 128x128 to 1024x1024. Large images which require using a disk-based pixel cache benefit from large tile sizes while images which fit in memory work well with smaller tile sizes.

tiff:tile-width=<width>

Specify the tile width to use while writing tiled TIFF. The tile height is then defaulted to an appropriate size. Width should be a multiple of 16. If the value is not a multiple of 16, then it will be rounded down. Enables tiled TIFF if it has not already been enabled.

tiff:tile-height=<height>

Specify the tile height to use while writing tiled TIFF. The tile width is then defaulted to an appropriate size. Height should be a multiple of 16. If the value is not a multiple of 16, then it will be rounded down. Enables tiled TIFF if it has not already been enabled.

tiff:webp-lossless={TRUE|FALSE}

Specify a value of TRUE to enable lossless mode while writing WebP-compressed TIFF files. The WebP webp:lossless option may also be used. The quality factor set by the -quality option may be used to influence the level of effort expended while compressing.

tiff:zstd-compress-level=<value>

Specify the compression level to use while writing Zstd-compressed TIFF files. The valid range is 1 to 22. If this define is not specified, then the 'quality' value is used such that the default quality setting of 75 is translated to a compress level of 9 such that 'quality' has a useful range of 10-184 if used for this purpose.

webp:exact={true|false}

Preserve exact RGB values under transparent areas. Default false.

webp:lossless={true|false}

Enable lossless encoding. Also enables 'exact' mode.

webp:method={0-6}

Quality/speed trade-off.

webp:image-hint={default,graph,photo,picture}

Hint for image type.

webp:target-size=<integer>

Target size in bytes.

webp:target-psnr=<float>

Minimal distortion to try to achieve.

webp:segments={1-4}

Maximum number of segments to use.

webp:sns-strength={0-100}

Spatial Noise Shaping.

webp:filter-strength={0-100}

Filter strength.

webp:filter-sharpness={0-7}

Filter sharpness.

webp:filter-type={0,1}

Filtering type. 0 = simple, 1 = strong (only used if filter-strength > 0 or autofilter is enabled).

webp:auto-filter={true|false}

Auto adjust filter's strength.

webp:alpha-compression=<integer>

Algorithm for encoding the alpha plane (0 = none, 1 = compressed with WebP lossless). Default is 1.

webp:alpha-filtering=<integer>

Predictive filtering method for alpha plane. 0: none, 1: fast, 2: best. Default is 1.

webp:alpha-quality={0-100}

Between 0 (smallest size) and 100 (lossless). Default is 100.

webp:pass=[1..10]

Number of entropy-analysis passes.

webp:show-compressed={true|false}

Export the compressed picture back. In-loop filtering is not applied.

webp:preprocessing=[0,1,2]

0=none, 1=segment-smooth, 2=pseudo-random dithering

webp:partitions=[0-3]

log2(number of token partitions) in [0..3]. Default is 0 for easier progressive decoding.

webp:partition-limit={0-100}

Quality degradation allowed to fit the 512k limit on prediction modes coding (0: no degradation, 100: maximum possible degradation).

webp:emulate-jpeg-size={true|false}

If true, compression parameters will be remapped to better match the expected output size from JPEG compression. Generally, the output size will be similar but the degradation will be lower.

webp:thread-level=<integer>

If non-zero, try and use multi-threaded encoding.

webp:low-memory={true|false}

If set, reduce memory usage (but increase CPU use)

webp:use-sharp-yuv={true|false}

If set, if needed, use sharp (and slow) RGB->YUV conversion

    gm convert bilevel.tif -define ps:imagemask eps3:stencil.ps

    Undefined       No disposal specified.
    None            Do not dispose between frames.
    Background      Overwrite the image area with
                    the background color.
    Previous        Overwrite the image area with
                    what was there prior to rendering
                    the image.

     point           x,y
     line            x0,y0 x1,y1
     rectangle       x0,y0 x1,y1
     roundRectangle  x0,y0 x1,y1 wc,hc
     arc             x0,y0 x1,y1 a0,a1
     ellipse         x0,y0 rx,ry a0,a1
     circle          x0,y0 x1,y1
     polyline        x0,y0  ...  xn,yn
     polygon         x0,y0  ...  xn,yn
     Bezier          x0,y0  ...  xn,yn
     path            path specification
     image           operator x0,y0 w,h filename

     text            x0,y0 string

     gravity         NorthWest, North, NorthEast, West, Center,
                     East, SouthWest, South, or SouthEast

     rotate          degrees
     translate       dx,dy
     scale           sx,sy
     skewX           degrees
     skewY           degrees

     color           x0,y0 method
     matte           x0,y0 method

     -draw 'circle 100,100 150,150'

     -draw 'image Over 100,100 225,225 image.jpg'

     point
     replace
     floodfill
     filltoborder
     reset

    gm convert infile.jpg -thumbnail 120x80 -background red -gravity center \
              -extent 140x100-10-5 outfile.jpg

    gm convert -size 800x600 input.jpg \
              -resize 800x600 -background black \
              -compose Copy -gravity center \
              -extent 800x600 \
              -quality 92 output.jpg

    name                              (named color)
    #RGB                              (hex numbers, 4 bits each)
    #RRGGBB                           (8 bits each)
    #RRRGGGBBB                        (12 bits each)
    #RRRRGGGGBBBB                     (16 bits each)
    #RRRRRRRRGGGGGGGGBBBBBBBB         (32 bits each)
    #RGBA                             (4 bits each)
    #RRGGBBAA                         (8 bits each)
    #RRRRRRGGGGGGBBBBBB               (12 bits each)
    #RRRRRRRRGGGGGGGGBBBBBBBBAAAAAAAA (32 bits each)
    rgb(r,g,b)                        (r,g,b are decimal numbers, range 0 to QuantumDepth)
    rgb(r%,g%,b%)                     (r,g,b are decimal numbers, range 0 to 100%)
    rgba(r,g,b,a)                     (r,g,b,a are decimal numbers, range 0 to QuantumDepth)
    rgba(r%,g%,b%,a%)                 (r,g,b,a are decimal numbers, range 0 to 100%)

    gm convert -fill blue ...
    gm convert -fill "#ddddff" ...
    gm convert -fill "rgb(65000,65000,65535)" ...

     Point
     Box
     Triangle
     Hermite
     Hanning
     Hamming
     Blackman
     Gaussian
     Quadratic
     Cubic
     Catrom
     Mitchell
     Lanczos
     Bessel
     Sinc

    gm convert -size 640x300 xc:transparent \
              -compose over -page +0-100 \
              frame.png -flatten output.png

    gm convert input.psd -flatten output.png

     %b   file size
     %c   comment
     %d   directory
     %e   filename extension
     %f   filename
     %g   page dimensions and offsets
     %h   height
     %i   input filename
     %k   number of unique colors
     %l   label
     %m   magick
     %n   number of scenes
     %o   output filename
     %p   page number
     %q   image bit depth
     %r   image type description
     %s   scene number
     %t   top of filename
     %u   unique temporary filename
     %w   width
     %x   horizontal resolution
     %y   vertical resolution
     %A   transparency supported
     %C   compression type
     %D   GIF disposal method
     %G   Original width and height
     %H   page height
     %M   original filename specification
     %O   page offset (x,y)
     %P   page dimensions (width,height)
     %Q   compression quality
     %T   time delay (in centi-seconds)
     %U   resolution units
     %W   page width
     %X   page horizontal offset (x)
     %Y   page vertical offset (y)
     %@   trim bounding box
     %#   signature
     \n   newline
     \r   carriage return
     %%   %

     -format "%m:%f %wx%h"

     Bilevel
     Grayscale
     GrayscaleMatte
     Palette
     PaletteMatte
     TrueColor
     TrueColorMatte
     ColorSeparation
     ColorSeparationMatte
     Optimize

     %[EXIF:<tag>]

     *  (print all Exif tags, in keyword=data format)
     !  (print all Exif tags, in tag_number format)
     #hhhh (print data for Exif tag #hhhh)
     ImageWidth
     ImageLength
     BitsPerSample
     Compression
     PhotometricInterpretation
     FillOrder
     DocumentName
     ImageDescription
     Make
     Model
     StripOffsets
     Orientation
     SamplesPerPixel
     RowsPerStrip
     StripByteCounts
     XResolution
     YResolution
     PlanarConfiguration
     ResolutionUnit
     TransferFunction
     Software
     DateTime
     Artist
     WhitePoint
     PrimaryChromaticities
     TransferRange
     JPEGProc
     JPEGInterchangeFormat
     JPEGInterchangeFormatLength
     YCbCrCoefficients
     YCbCrSubSampling
     YCbCrPositioning
     ReferenceBlackWhite
     CFARepeatPatternDim
     CFAPattern
     BatteryLevel
     Copyright
     ExposureTime
     FNumber
     IPTC/NAA
     ExifOffset
     InterColorProfile
     ExposureProgram
     SpectralSensitivity
     GPSInfo
     ISOSpeedRatings
     OECF
     ExifVersion
     DateTimeOriginal
     DateTimeDigitized
     ComponentsConfiguration
     CompressedBitsPerPixel
     ShutterSpeedValue
     ApertureValue
     BrightnessValue
     ExposureBiasValue
     MaxApertureValue
     SubjectDistance
     MeteringMode
     LightSource
     Flash
     FocalLength
     MakerNote
     UserComment
     SubSecTime
     SubSecTimeOriginal
     SubSecTimeDigitized
     FlashPixVersion
     ColorSpace
     ExifImageWidth
     ExifImageLength
     InteroperabilityOffset
     FlashEnergy
     SpatialFrequencyResponse
     FocalPlaneXResolution
     FocalPlaneYResolution
     FocalPlaneResolutionUnit
     SubjectLocation
     ExposureIndex
     SensingMethod
     FileSource
     SceneType

     %[JPEG-<tag>]

     *                 (all JPEG-related tags, in
                        keyword=data format)
     Quality           IJG JPEG "quality" estimate
     Colorspace        JPEG colorspace numeric ID
     Colorspace-Name   JPEG colorspace name
     Sampling-factors  JPEG sampling factors

     -label "%m:%f %wx%h"

     -colorspace gray -lat "10x10-5%"

     %d   domain
     %e   event
     %f   function
     %l   line
     %m   module
     %p   process ID
     %r   real CPU time
     %t   wall clock time
     %u   user CPU time
     %%   percent sign
     \n   newline
     \r   carriage return

    gm convert -debug coders -log "%u %m:%l %e" in.gif out.png

     best
     default
     gray
     red
     green
     blue

    gm convert -background black \
              -compose CopyRed   -page +0-100 red.png \
              -compose CopyGreen -page +0+40  green.png \
              -compose CopyBlue  -page +0+180 blue.png \
              -mosaic output.png

     Uniform
     Gaussian
     Multiplicative
     Impulse
     Laplacian
     Poisson
     Random (uniform distribution)

Add

Result is rvalue added to channel value.

And

Result is the logical AND of rvalue with channel value.

Assign

Result is rvalue.

Depth

Result is channel value adjusted so that it may be (approximately) stored in the specified number of bits without additional loss.

Divide

Result is channel value divided by rvalue.

Gamma

Result is channel value gamma adjusted by rvalue.

LShift

Result is channel value bitwise left shifted by rvalue bits.

Log

Result is computed as log(value*rvalue+1)/log(rvalue+1).

Max

Result is assigned to rvalue if rvalue is greater than value.

Min

Result is assigned to rvalue if rvalue is less than value.

Multiply

Result is channel value multiplied by rvalue.

Negate

Result is inverse of channel value (like a film negative). An rvalue must be supplied but is currently not used. Inverting the image twice results in the original image.

Or

Result is the logical OR of rvalue with channel value.

Pow

Result is computed as pow(value,rvalue). Similar to Gamma except that rvalue is not inverted.

RShift

Result is channel value bitwise right shifted by rvalue bits.

Subtract

Result is channel value minus rvalue.

Threshold

Result is maximum (white) if channel value is greater than rvalue, or minimum (black) if it is less than or equal to rvalue. If all channels are specified, then thresholding is done based on computed pixel intensity.

Threshold-white

Result is maximum (white) if channel value is greater than rvalue and is unchanged if it is less than or equal to rvalue. This can be used to remove apparent noise from the bright parts of an image. If all channels are specified, then thresholding is done based on computed pixel intensity.

Threshold-White-Negate

Result is set to black if channel value is greater than rvalue and is unchanged if it is less than or equal to rvalue. If all channels are specified, then thresholding is done based on computed pixel intensity.

Threshold-black

Result is minimum (black) if channel value is less than than rvalue and is unchanged if it is greater than or equal to rvalue. This can be used to remove apparent noise from the dark parts of an image. If all channels are specified, then thresholding is done based on computed pixel intensity.

Threshold-Black-Negate

Result is set to white if channel value is less than than rvalue and is unchanged if it is greater than or equal to rvalue. If all channels are specified, then thresholding is done based on computed pixel intensity.

Xor

Result is the logical XOR of rvalue with channel value. An interesting property of XOR is that performing the same operation twice results in the original value.

Noise-Gaussian

Result is the current channel value modulated with gaussian noise according to the intensity specified by rvalue.

Noise-Impulse

Result is the current channel value modulated with impulse noise according to the intensity specified by rvalue.

Noise-Laplacian

Result is the current channel value modulated with laplacian noise according to the intensity specified by rvalue.

Noise-Multiplicative

Result is the current channel value modulated with multiplicative gaussian noise according to the intensity specified by rvalue.

Noise-Poisson

Result is the current channel value modulated with poisson noise according to the intensity specified by rvalue.

Noise-Random

Result is the current channel value modulated with random (uniform distribution) noise according to the intensity specified by rvalue. The initial noise intensity (rvalue=1.0) is the range of one pixel quantum span.

Noise-Uniform

Result is the channel value with uniform noise applied according to the intensity specified by rvalue.

    gm convert in.bmp -operator red assign "50%" out.bmp

    gm convert in.bmp -operator gray threshold "50%" out.bmp

     11x17         792  1224
     Ledger       1224   792
     Legal         612  1008
     Letter        612   792
     LetterSmall   612   792
     ArchE        2592  3456
     ArchD        1728  2592
     ArchC        1296  1728
     ArchB         864  1296
     ArchA         648   864
     A0           2380  3368
     A1           1684  2380
     A2           1190  1684
     A3            842  1190
     A4            595   842
     A4Small       595   842
     A5            421   595
     A6            297   421
     A7            210   297
     A8            148   210
     A9            105   148
     A10            74   105
     B0           2836  4008
     B1           2004  2836
     B2           1418  2004
     B3           1002  1418
     B4            709  1002
     B5            501   709
     C0           2600  3677
     C1           1837  2600
     C2           1298  1837
     C3            918  1298
     C4            649   918
     C5            459   649
     C6            323   459
     Flsa          612   936
     Flse          612   936
     HalfLetter    396   612

     Rotate
     Shear
     Roll
     Hue
     Saturation
     Brightness
     Gamma
     Spiff
     Dull
     Grayscale
     Quantize
     Despeckle
     ReduceNoise
     AddNoise
     Sharpen
     Blur
     Threshold
     EdgeDetect
     Spread
     Shade
     Raise
     Segment
     Solarize
     Swirl
     Implode
     Wave
     OilPaint
     CharcoalDrawing
     JPEG

    gm convert cockatoo.jpg exifdata.app1

     0: none
     1: sub
     2: up
     3: average
     4: Paeth

  1 0 0
  0 1 0
  0 0 1

  -recolor "1 0 0, 0 1 0, 0 0 1"

  1 0 0 0
  0 1 0 0
  0 0 1 0
  0 0 0 1

  1 0 0 0 0
  0 1 0 0 0
  0 0 1 0 0
  0 0 0 1 0
  0 0 0 0 1

  0 0 1
  0 1 0
  1 0 0

  0.2126 0.7152 0.0722
  0.2126 0.7152 0.0722
  0.2126 0.7152 0.0722

  0.8 0.0 0.0 0.0 0.1
  0.0 0.8 0.0 0.0 0.1
  0.0 0.0 0.8 0.0 0.1
  0.0 0.0 0.0 0.8 0.1
  0.0 0.0 0.0 0.0 1.0