Add Hough transform (#32)

* Ignore IntelliJ metadata files

* Add Hough line transform

Signed-off-by: Gabe Selzer <[email protected]>

* HoughLineTransformTest: Optimize performance

* HoughLineTransformTest: add missing javadoc

* HoughLineTransform: Clone peak storage array

This stage of the transform makes use of a single array to grab each
peak Point and then store it in an ArrayList of long[]. However the
array was not copied as it was added into the ArrayList, resulting in
all entries in the ArrayList pointing to the same array. This commit
fixes this bug through use of Arrays.copyOf() to copy each array and add
the result to the ArrayList.

* HoughLineTransform: fix rawType warning

* HoughTransform: remove Benchmark import

This import is not necessary for the class since the class is not a
benchmark.

* HoughLineTransform: Refactor into static methods.

The previous implementation was based on stateful objects, which was not
optimal for use of this algorithm. The algorithm was thus reworked into
static methods for both votespace generation and peak picking. The
process of the algorithm was kept but massaged to fit the new
implementation style.

* PickPeaks: Allow user input for min peak value.

* PickPeaks: Translate votespace for correct output

PickPeaks now returns a rho value between [-maxRho, maxRho), whereas
before it returned a rho value between [0, maxRho * 2). Translating the
votespace before finding the peaks allows for easier manipulation of the
polar coordinates, and prevents the user from having to translate the
coordinates or votespace themselves.

* LineTest: Add PickPeaks regression test

* HoughTransforms: Add slope/intercept converters

These methods convert the polar output of PickPeaks into more useful
slope and intercept values for lines. In the case of an undefined slope
getSlope() returns Double.POSITIVE_INFINITY or DOUBLE.NEGATIVE_INFINITY

* Cleanup / format

* Improve line transform documentation

Javadoc adapted from previous commit c1952b1, HoughTransforms.java, line 50

* computeLength: allow larger return values

* defaultRho: simplify instructions to one-liner

* pickPeaks: loosen typing of parameters

* voteLines: Restrict voteSpace to only IntegerType

Since the values in the votespace are only ever incremented by one in
this particular transform we restrict the votespace to
RandomAccessibleIntervals of types extending IntegerType only

* pickPeaks: add alternative (convenient) option

* voteLines: Relax typing of input RAI

* fixup! pickPeaks: add alternative (convenient) option

* voteLines: Remove unnecessary variables

* HoughLineTest: use small delta for slope and yint

* HoughTransforms: Loosen helper method parameters

The getVotespaceSize() and defaultRho() method parameter was loosened
from being a RandomAccessibleInterval to being a Dimensions object to
allow a wider range of acceptible input types

* HoughLine: Remove unnecessary array

* PickPeaks: Fix translation bug

* PickPeaks: Rename to PickLinePeaks

Since we are translating the vote space before picking peaks (something
that future Hough Transforms might not do), we will change the name to
something more specific to the line transform.

* VoteLines: Refactor parameter order

This allows a theta-only (i.e. to allow a theta but not a rho to be
    passed through) signature to make sense and to keep the same
parameter order

* VoteLines: Move check outside for-loop

* HoughTransforms: Remove unused import

* HoughLineTest: Refactor to better input image

* getVotespaceSize: Allow for non-default thetas

* HoughTransforms: Formatting

* VoteLines: Translate image for zeroed min

* VoteLines: Refactor LUTs

this commit removes the unused rho[] LUT and refactors the theta[] LUT
into two different LUTs for sin(theta) and cos(theta). This refactor
speeds up later calculations

* VoteLines: fix predicate bug

This commit changes the Predicate deciding whether to allow votespace
population from being a strictly greater-than to a
greater-than-or-equal-to comparison. This fix was made to prevent a bug
in which no threshold is provided to the algorithm and at the same time
the max value of the image is at the first index of the image. If the
comparison is not <= then all pixels will fail the Predicate, since no
pixel is greater than the max. This means that passing a value to the
algorithm is still preferable but allows more scenarios where no value
is given.

* VoteLines: Refactor fRho/r declaration

* VoteLines: Remove cursor positions from for loop

* VoteLines: remove unnecessary float cast.

* HoughLineTest: use ArrayImgs to make votespace

* HoughLineTest: Formatting

* VoteLines: Fix javadoc

Author: panovr

.gitignore

@@ -2,3 +2,7 @@
 /.project
 /.settings/
 /target/
+
+# IntelliJ
+/.idea/
+*.iml