qa.qcri.prepocressor.imageprocessing

Class Geometry

java.lang.Object

qa.qcri.prepocressor.imageprocessing.Geometry

public final class Geometry
extends java.lang.Object

Provides geometric functions.

Author:

Felix Stahlberg

Method Summary

Methods

Modifier and Type	Method and Description
static org.opencv.core.Mat	componentDensity(org.opencv.core.Mat input, int minSize) Calculates a map of connected component density.
static org.opencv.core.Mat	dct(org.opencv.core.Mat input) Discrete Cosine transform.
static org.opencv.core.Mat	dft(org.opencv.core.Mat input) Discrete Fourier transform.
static org.opencv.core.Mat	myMedianBlur(org.opencv.core.Mat input, int xSize, int ySize) OpenCV does not support median filtering with non-quadratic kernels.
static org.opencv.core.Mat	rectSum(org.opencv.core.Mat input, int fixRow, int fixCol, int maxHeight, int maxWidth) Calculates the sum of elements within rectangles in the image.
static org.opencv.core.Mat	reducedALCM(org.opencv.core.Mat input, org.opencv.core.Size kSize, int resolution) Applies a steerable ellipsoid filter to create an adaptive local connectivity map.
static org.opencv.core.Mat	reducedRunlength(org.opencv.core.Mat input, int resolution) Computes a connectivity map based on run length from pixels in specific directions.
static org.opencv.core.Mat	removeLargeComponents(org.opencv.core.Mat input, int maxWidth, int maxHeight) Remove large connected components in the image.
static org.opencv.core.Mat	runlengthMap(org.opencv.core.Mat input, double theta)
static org.opencv.core.Mat	thinning(org.opencv.core.Mat im) Function for thinning the given binary image

Methods inherited from class java.lang.Object

equals, getClass, hashCode, notify, notifyAll, toString, wait, wait, wait

Method Detail

rectSum

public static org.opencv.core.Mat rectSum(org.opencv.core.Mat input,
                          int fixRow,
                          int fixCol,
                          int maxHeight,
                          int maxWidth)

Calculates the sum of elements within rectangles in the image. The rectangles have a common edge point (fix point) but their width and height vary. Produces a maxHeight times maxWidth matrix storing the sum within corresponding rectangles.

Parameters:

input - input image

fixRow - row number of the fix point

fixCol - column number of the fix point

maxHeight - maximum height of the rectangles

maxWidth - maximum width of the rectangles

Returns:

maxHeight times maxWidth matrix with rectangle sums

componentDensity

public static org.opencv.core.Mat componentDensity(org.opencv.core.Mat input,
                                   int minSize)

Calculates a map of connected component density. Each connected component adds 1/area to all pixels in its bounding box where area is the size of the bounding box.

Parameters:

input - input image

minSize - bounding boxes below this size are ignored

Returns:

density map

removeLargeComponents

public static org.opencv.core.Mat removeLargeComponents(org.opencv.core.Mat input,
                                        int maxWidth,
                                        int maxHeight)

Remove large connected components in the image.

Parameters:

input - input image

maxWidth - maximum width of connected components

maxHeight - maximum height of connected components

Returns:

input image without large connected components

dft

public static org.opencv.core.Mat dft(org.opencv.core.Mat input)

Discrete Fourier transform.

Parameters:

input - input image

Returns:

Fourier transformed image

dct

public static org.opencv.core.Mat dct(org.opencv.core.Mat input)

Discrete Cosine transform.

Parameters:

input - input image

Returns:

Cosine transformed image

thinning

public static org.opencv.core.Mat thinning(org.opencv.core.Mat im)

Function for thinning the given binary image

Parameters:

im - Binary image with range = 0-255

See Also:

http://opencv-code.com/quick-tips/implementation-of-thinning-algorithm-in-opencv/

reducedALCM

public static org.opencv.core.Mat reducedALCM(org.opencv.core.Mat input,
                              org.opencv.core.Size kSize,
                              int resolution)
                                       throws java.lang.IllegalArgumentException

Applies a steerable ellipsoid filter to create an adaptive local connectivity map. The ALCM of each direction is reduced horizontally to a single col. The i-th col of the resulting image corresponds to the direction i. i encodes the angle 180*i/resolution. If resolution=2, i=0 is the horizontal, i=1 the vertical ALCM. The resulting image has the width resolution.

Parameters:

input - input image

kSize - size of the kernel. Should have width>height

resolution - angle resolution and width of the output image. >0.

Returns:

Matrix consisting of cols corresponding to horizontally reduced ALCMs in a specific direction

Throws:

java.lang.IllegalArgumentException - if resolution < 1 or height > width

java.lang.NullPointerException - if input is null

reducedRunlength

public static org.opencv.core.Mat reducedRunlength(org.opencv.core.Mat input,
                                   int resolution)
                                            throws java.lang.IllegalArgumentException

Computes a connectivity map based on run length from pixels in specific directions. The i-th col of the resulting image corresponds to the direction i. i encodes the angle 180*i/resolution. If resolution=2, i=0 is the horizontal, i=1 the vertical ALCM. The resulting image has the width resolution.

Parameters:

input - input image

resolution - angle resolution and width of the output image. >0.

Returns:

Matrix consisting of cols corresponding to horizontally reduced run length information

Throws:

java.lang.IllegalArgumentException - if resolution < 1

java.lang.NullPointerException - if input is null

runlengthMap

public static org.opencv.core.Mat runlengthMap(org.opencv.core.Mat input,
                               double theta)

myMedianBlur

public static org.opencv.core.Mat myMedianBlur(org.opencv.core.Mat input,
                               int xSize,
                               int ySize)

OpenCV does not support median filtering with non-quadratic kernels. Therefore, this is a slower but more flexible medianBlur implementation

Parameters:

input - input image

xSize - kernel size in x direction

ySize - kernel size in y direction

Returns:

blurred image