ConformImages

ConformImages[{image1,image2,}]

returns a list of images where all imagei are made to have conforming properties, including dimensions, data type, color space, and interleaving.

ConformImages[{image1,image2,},spec]

returns all images of the specified spec.

ConformImages[{image1,image2,},spec,fitting]

resizes images using the specified fitting method.

Details and Options

  • ConformImages is typically used as a preprocessing step to algorithms that need images that are uniform in some way, such as classification, dimensionality reduction, and assembling into bigger images.
  • ConformImages works with a list of 2D as well as 3D images.
  • Image specification spec can be one of the following:
  • imagetake the conforming properties from image
    sizeimage dimensions specification
    imagetypeimage type as in ImageType
    {spec1,spec2,}specification for multiple properties
  • By default, the smallest imagetype that can represent the data stored in all imagei is used.
  • The following size specifications can be used:
  • wwidth w, using a height to give the median aspect ratio
    {w,h}width and height of the output image
    {w,d,h}width, depth, and height of a 3D image
  • Any of the w, d, and h can be one of the following:
  • valscalar value
    Tiny,Small,Medium,Largepredefined absolute sizes
    ffunction f to be applied to a list of sizes in each dimension
  • ConformImages[{image1,image2,}] resizes all images to have the median width and a height to give the median aspect ratio.
  • Possible values for fitting include:
  • "Fill"fit the smaller dimension; crop the other if necessary
    "Fit"fit the whole image; keep the aspect ratio; pad if necessary
    "Pad"pad and/or crop without resampling
    "Stretch"stretch the image to fit by resampling
    "Letterbox"fit the width; pad and/or crop in other dimensions
    "Pillarbox"fit the height; pad and/or crop in other dimensions
    "Windowbox"fit to the center; pad all sides
  • The default fitting method is "Stretch".
  • ConformImages has the same options as Image and Image3D with the following additions and changes:
  • ColorSpaceAutomaticoutput color space
    ImageSizeAutomaticthe default displayed size of the image
    InterleavingAutomaticwhether to interleave channels
    MagnificationAutomatichow to magnify the displayed image
    PaddingAutomaticmethod used to make same-size images
  • By default, all other option settings are inherited from individual input images, which would be equivalent to option->Inherited.
  • With ColorSpace->Automatic, a color space in which all images can be represented is chosen.
  • With Interleaving->Automatic, interleaving is set to False only if no input image is interleaved.

Examples

open allclose all

Basic Examples  (2)

Conform images:

Conform 3D images:

Scope  (6)

Image Size  (1)

Conform image sizes using automatic sizing:

Conform to a specific size:

Conform each image to have the maximum width and automatic height:

Use smallest width and height:

Use smallest width and largest height:

Image Properties  (3)

Specify the common data type:

Specify both size and data type:

Use an image to specify conforming properties:

Fitting Methods  (2)

By default, images are stretched to fit in the common size:

Specify a different fitting method:

Conform properties of a list of 3D images:

Options  (6)

Boxed  (1)

For 3D images, the Boxed option is inherited from original images:

Use Boxed->True for all images:

ColorFunction  (2)

By default, ColorFunction is set to Automatic:

Specify the output color function:

ColorFunction is set to Automatic if the input and output color spaces do not match:

ColorSpace  (1)

By default, a color space that can store the most image information is selected:

Specify the output color space:

Padding  (2)

Default padding value is chosen based on the fitting method:

With "Pad" fitting method, a common border color is used when available. Otherwise black is used:

Specify the padding value:

Black padding is used with "Letterbox", "Pillarbox", and "Windowbox" fitting methods:

Applications  (2)

Conform a list of face images:

Assemble an array of images:

Properties & Relations  (1)

For multichannel images with a different number of channels, the maximum number of channels is used:

Zero is used for non-existing channels:

Possible Issues  (1)

For images with ColorSpace->Automatic, pixel values with fewer channels are zero padded:

Specify the output ColorSpace:

Introduced in 2014
 (10.0)