Package org.apache.sis.image

Class ResampledImage

java.lang.Object
org.apache.sis.image.PlanarImage
org.apache.sis.image.ComputedImage
org.apache.sis.image.ResampledImage
All Implemented Interfaces:
RenderedImage, Disposable
Direct Known Subclasses:
Visualization
public class ResampledImage extends ComputedImage
An image which is the result of resampling the pixel values of another image. Resampling is the action of computing pixel values at possibly non-integral positions of a source image. It can be used for projecting an image to another coordinate reference system, for example from (latitude, longitude) to World Mercator. The resampling is defined by a non-linear MathTransform (for example a map projection) which converts pixel center coordinates from this image to pixel center coordinates in the source image. The converted coordinates usually contain fraction digits, in which case an interpolation is applied.

Usage note

This class should be used with non-linear transforms such as map projections. It is technically possible to use this class with linear transforms such as AffineTransform, but there is more efficient alternatives for linear cases (for example specifying the affine transform at rendering time).
Since:
1.1
Version:
1.2
See Also:

  • Field Details

    • POSITIONAL_CONSISTENCY_KEY

      public static final String POSITIONAL_CONSISTENCY_KEY
      Key of a property providing an estimation of positional error for each pixel. Values shall be instances of RenderedImage with same size and origin than this image. The image should contain a single band where all sample values are error estimations in pixel units (relative to pixels of this image). The value should be small, for example between 0 and 0.2.

      The default implementation transforms all pixel coordinates to source, then convert them back to pixel coordinates in this image. The result is compared with expected coordinates and the distance is stored in the image.

      See Also:

    • BIDIMENSIONAL

      static final int BIDIMENSIONAL
      The 2 value for identifying code expecting exactly 2 dimensions.
      See Also:

    • minX

      private final int minX
      Domain of pixel coordinates in this image.

    • minY

      private final int minY
      Domain of pixel coordinates in this image.

    • width

      private final int width
      Domain of pixel coordinates in this image.

    • height

      private final int height
      Domain of pixel coordinates in this image.

    • minTileX

      private final int minTileX
      Index of the first tile.

    • minTileY

      private final int minTileY
      Index of the first tile.

    • toSource

      protected final org.opengis.referencing.operation.MathTransform toSource
      Conversion from pixel center coordinates of this image to pixel center coordinates of source image. This transform should be an instance of MathTransform2D, but this is not required by this class (a future version may allow interpolations in a n-dimensional cube).
      See Also:
      • PixelInCell.CELL_CENTER

    • toSourceSupport

      private final org.opengis.referencing.operation.MathTransform toSourceSupport
      Same as toSource but with the addition of a shift for taking in account the number of pixels required for interpolations. For example if a bicubic interpolation needs 4×4 pixels, then the source coordinates that we need are not the coordinates of the pixel we want to interpolate, but 1 or 2 pixels before for making room for interpolation support.

      This transform may be an instance of ResamplingGrid if the usage of such grid has been authorized. That transform may be non-invertible. Consequently, this transform should not be used for inverse operations and should not be made accessible to the user.

      This transform maps pixel centers of both images, except in the case of nearest-neighbor interpolation. In that special case only, the transform maps target pixel center to source pixel corner. We have to map corners in source images because source pixel coordinates are computed by taking the integer parts of toSourceSupport results, without rounding.

      See Also:

    • interpolation

      protected final Interpolation interpolation
      The object to use for performing interpolations.

    • fillValues

      private final Object fillValues
      The values to use if a pixel in this image cannot be mapped to a pixel in the source image. Must be an int[] or double[] array (no other type allowed). The array length must be equal to the number of bands. Cannot be null.

    • linearAccuracy

      private final javax.measure.Quantity<javax.measure.quantity.Length> linearAccuracy
      The largest accuracy declared in the accuracy argument given to constructor, or null if none. This is for information purpose only.
      See Also:

    • positionalConsistency

      private Reference<ComputedImage> positionalConsistency
      POSITIONAL_CONSISTENCY_KEY value, computed when first requested.
      See Also:

    • mask

      private Reference<ComputedImage> mask
      PlanarImage.MASK_KEY value, computed when first requested.
      See Also:

  • Constructor Details

    • ResampledImage

      protected ResampledImage(RenderedImage source, SampleModel sampleModel, Point minTile, Rectangle bounds, org.opengis.referencing.operation.MathTransform toSource, Interpolation interpolation, Number[] fillValues, javax.measure.Quantity<?>[] accuracy)
      Creates a new image which will resample the given image. The resampling operation is defined by a potentially non-linear transform from this image to the specified source image. That transform should map pixel centers.

      The sampleModel determines the tile size and the target data type. This is often the same sample model than the one used by the source image, but may also be different for forcing a different tile size or a different data type (e.g. byte versus float) for storing resampled values. If the specified sample model is not the same than the one used by the source image, then subclass should override getColorModel() for returning a color model which is compatible with the sample model.

      If a pixel in this image cannot be mapped to a pixel in the source image, then the sample values are set to fillValues. If the given array is null, or if any element in the given array is null, then the default fill value is NaN for floating point data types or zero for integer data types. If the array is shorter than the number of bands, then above-cited default values are used for missing values. If longer than the number of bands, extraneous values are ignored.

      Parameters:
      source - the image to be resampled.
      sampleModel - the sample model shared by all tiles in this resampled image.
      minTile - indices of the first tile (minTileX, minTileY), or null for (0,0).
      bounds - domain of pixel coordinates of this resampled image.
      toSource - conversion of pixel coordinates of this image to pixel coordinates of source image.
      interpolation - the object to use for performing interpolations.
      fillValues - the values to use for pixels in this image that cannot be mapped to pixels in source image. May be null or contain null elements, and may have any length (see above for more details).
      accuracy - values of "org.apache.sis.PositionalAccuracy" property, or null if none. This constructor may retain only a subset of specified values or replace some of them. If an accuracy is specified in pixel units, then a value such as 0.125 pixel may enable the use of a slightly faster algorithm at the expense of accuracy. This is only a hint honored on a best-effort basis.
      See Also:

  • Method Details

    • interpolationSupportOffset

      static double interpolationSupportOffset(int span)
      The relative index of the first pixel needed on the left or top side of the region to use for interpolation. The index is relative to the two central pixels needed for a bilinear interpolation. This value is negative or zero. A value of 0 means that we need no pixels in addition of the two central pixels:
      sample[0] … (position where to interpolate) … sample[1]
      A value of -1 means that we need one more pixel on the left or top side. It often means that we also need one more pixel on the right or bottom sides, but not necessarily; those sides are not this method business.
      sample[-1] … sample[0] … (position where to interpolate) … sample[1] … sample[2]

      Nearest-neighbor special case

      The nearest-neighbor interpolation (identified by span == 1) is handled in a special way. The return value should be 0 according above contract, but this method returns 0.5 instead. This addition of a 0.5 offset allows the following substitution: Math.round(double) is the desired behavior for nearest-neighbor interpolation, but the buffer given to Interpolation.interpolate(DoubleBuffer, int, double, double, double[], int) is filled with values at coordinates determined by Math.floor(double) semantic. Because the buffer has only one value, interpolate(…) has no way to look at neighbor values for the best match (contrarily to what other interpolation implicitly do, through mathematics). The 0.5 offset is necessary for compensating.
      Parameters:
      span - the width or height of the support region for interpolations.
      Returns:
      relative index of the first pixel needed on the left or top sides, as a value ≤ 0 (except in nearest-neighbor special case).
      See Also:

    • interpolationLimit

      private static double interpolationLimit(double max, int span)
      Returns the upper limit (inclusive) where an interpolation is possible. The given max value is the maximal coordinate value (inclusive) traversed by PixelIterator. Note that this is not the image size because of margin required by interpolation methods.

      Since interpolator will receive data at coordinates max to max + span - 1 inclusive and since those coordinates are pixel centers, the points to interpolate are on the surface of a valid pixel until (max + span - 1) + 0.5. Consequently, this method computes max + span - 0.5. An additional 0.5 offset is added in the special case of nearest-neighbor interpolation for consistency with interpolationSupportOffset(int).

      Parameters:
      max - the maximal coordinate value, inclusive.
      span - the width or height of the support region for interpolations.
      Returns:
      max + span - 0.5 (except in nearest-neighbor special case).
      See Also:

    • getPositionalAccuracyCount

      private int getPositionalAccuracyCount()
      Returns the number of quantities in the array returned by getPositionalAccuracy().

    • getPositionalAccuracy

      private javax.measure.Quantity<?>[] getPositionalAccuracy()
      Computes the "org.apache.sis.PositionalAccuracy" value. This method is invoked by getProperty(String) when the PlanarImage.POSITIONAL_ACCURACY_KEY property value is requested.

    • getPositionalConsistency

      private RenderedImage getPositionalConsistency() throws org.opengis.referencing.operation.TransformException
      Computes the "org.apache.sis.PositionalConsistency" value. This method is invoked by getProperty(String) when the POSITIONAL_CONSISTENCY_KEY property value is requested. The result is saved by weak reference since recomputing this image is rarely requested, and if needed can be recomputed easily.
      Throws:
      org.opengis.referencing.operation.TransformException

    • getMask

      private RenderedImage getMask()
      Computes the "org.apache.sis.Mask" value. This method is invoked by getProperty(String) when the PlanarImage.MASK_KEY property value is requested. The result is saved by weak reference since recomputing this image is rarely requested, and if needed can be recomputed easily.

    • hasNoMask

      boolean hasNoMask()
      Returns true if this image cannot have mask.

    • verify

      public String verify()
      Verifies whether image layout information are consistent. This method verifies that source coordinates required by this image (computed by converting this image bounds using the toSource transform) intersects the bounds of the source image. If this is not the case, then this method returns "toSource" for signaling that the transform may have a problem. Otherwise this method completes the check with all verifications documented in parent class
      Overrides:
      verify in class PlanarImage
      Returns:
      null if image layout information are consistent, or the name of inconsistent attribute if a problem is found.

    • recoverableException

      private static void recoverableException(String method, Exception error)
      Invoked when a non-fatal error occurred.
      Parameters:
      method - the method where the ignorable error occurred.
      error - the ignore which can be ignored.

    • getColorModel

      public ColorModel getColorModel()
      Returns the color model of this resampled image. Default implementation assumes that this image has the same color model than the source image.
      Returns:
      the color model, or null if unspecified.

    • getProperty

      public Object getProperty(String key)
      Gets a property from this image. Current default implementation supports the following keys (more properties may be added to this list in future Apache SIS versions):
      Note: the sample resolutions are retained because they should have approximately the same values before and after resampling. Statistics are not in this list because, while minimum and maximum values should stay approximately the same, the average value and standard deviation may be quite different.
      Specified by:
      getProperty in interface RenderedImage
      Overrides:
      getProperty in class PlanarImage
      Parameters:
      key - the name of the property to get.
      Returns:
      the property value, or Image.UndefinedProperty if none.

    • getPropertyNames

      public String[] getPropertyNames()
      Returns the names of all recognized properties, or null if this image has no properties. The returned array contains the properties listed in getProperty(String) if the source image has those properties.
      Specified by:
      getPropertyNames in interface RenderedImage
      Overrides:
      getPropertyNames in class PlanarImage
      Returns:
      names of all recognized properties, or null if none.

    • getMinTileX

      public final int getMinTileX()
      Returns the minimum tile index in the x direction. This is often 0.
      Specified by:
      getMinTileX in interface RenderedImage
      Overrides:
      getMinTileX in class PlanarImage
      Returns:
      the minimum tile index in the x direction.

    • getMinTileY

      public final int getMinTileY()
      Returns the minimum tile index in the y direction. This is often 0.
      Specified by:
      getMinTileY in interface RenderedImage
      Overrides:
      getMinTileY in class PlanarImage
      Returns:
      the minimum tile index in the y direction.

    • getMinX

      public final int getMinX()
      Returns the minimum x coordinate (inclusive) of this image. This is the Rectangle.x value of the bounds specified at construction time.
      Specified by:
      getMinX in interface RenderedImage
      Overrides:
      getMinX in class PlanarImage
      Returns:
      the minimum x coordinate (column) of this image.

    • getMinY

      public final int getMinY()
      Returns the minimum y coordinate (inclusive) of this image. This is the Rectangle.y value of the bounds specified at construction time.
      Specified by:
      getMinY in interface RenderedImage
      Overrides:
      getMinY in class PlanarImage
      Returns:
      the minimum y coordinate (row) of this image.

    • getWidth

      public final int getWidth()
      Returns the number of columns in this image. This is the Rectangle.width value of the bounds specified at construction time.
      Returns:
      number of columns in this image.

    • getHeight

      public final int getHeight()
      Returns the number of rows in this image. This is the Rectangle.height value of the bounds specified at construction time.
      Returns:
      number of rows in this image.

    • computeTile

      protected Raster computeTile(int tileX, int tileY, WritableRaster tile) throws org.opengis.referencing.operation.TransformException
      Invoked when a tile need to be computed or updated. This method fills all pixel values of the tile with values interpolated from the source image. It may be invoked concurrently in different threads.
      Specified by:
      computeTile in class ComputedImage
      Parameters:
      tileX - the column index of the tile to compute.
      tileY - the row index of the tile to compute.
      tile - if the tile already exists but needs to be updated, the tile to update. Otherwise null.
      Returns:
      computed tile for the given indices.
      Throws:
      org.opengis.referencing.operation.TransformException - if an error occurred while computing pixel coordinates.

    • prefetch

      protected Disposable prefetch(Rectangle tiles)
      Notifies the source image that tiles will be computed soon in the given region. If the source image is an instance of ComputedImage, then this method forwards the notification to it. Otherwise default implementation does nothing.
      Overrides:
      prefetch in class ComputedImage
      Parameters:
      tiles - indices of the tiles which will be prefetched.
      Returns:
      handler on which to invoke dispose() after the prefetch operation completed (successfully or not), or null if none.
      Since:
      1.2

    • equals

      public boolean equals(Object object)
      Compares the given object with this image for equality. This method returns true if the given object is non-null, is an instance of the exact same class than this image, has equal sources and do the same resampling operation (same interpolation method, same fill values, same coordinates).
      Overrides:
      equals in class Object
      Parameters:
      object - the object to compare with this image.
      Returns:
      true if the given object is an image performing the same resampling than this image.

    • hashCode

      public int hashCode()
      Returns a hash code value for this image.
      Overrides:
      hashCode in class Object
      Returns:
      a hash code value based on a description of the operation performed by this image.