BlockProcessing

BlockProcessing[f, img, {n₁,...}] partitions img into non-overlapping blocks of size {n₁, ...} and applies function f to each block.

BlockProcessing[f, img, {n₁,...}, {d₁, ...}] partitions img into blocks of size {n₁, ...} with offset {d₁, ...} and applies function f to each block.

BlockProcessing will return a result of the same dimensions as img if function f returns a scalar.

If img is an ImageData expression, BlockProcessing will return an ImageData object only if the operation f returns a raw data structure that is ImageData compatible (see Equations (2.2.1), ( 2.3.1), (2.3.2)).

The option PaddingMethod may be used to select the type of border extension used during calculation. Default setting is PaddingMethod→FixedPad. Any of the supported padding methods may be used (ConstantPad, FixedPad, PeriodicPad, ReflectPad, ZeroPad). The setting PaddingMethod→None may be used to compute the result without extending the border.

The option Centered may be used to extend the border symmetrically or only at the end of a list. Default setting is Centered→False.

The argument img may be an array or ImageData expression.

See also User's Guide 5.4.

Modified in Version 2.

Examples

This loads the package.

In[1]:=

Here are some simple usage examples.

In[2]:=

This wraps function f around each 2 × 2 block of array tmp.

In[3]:=

Out[3]//MatrixForm=

This is an example of block processing with a sliding window.

In[4]:=

Out[4]//MatrixForm=

This disables border extension.

In[5]:=

Out[5]//MatrixForm=

Here we make function f return its argument by using the pure function #&.

In[6]:=

Out[6]//MatrixForm=

This joins the elements of all the submatrices.

In[7]:=

Out[7]//MatrixForm=

Here we use BlockProcessing to return the variance within 8×8 partially overlapping blocks in an example image.

In[8]:=

This shows the result.

In[9]:=

© 2008 Wolfram Research, Inc. • Terms of Use • Privacy Policy