Reference Guide¶
Table of Contents¶
polarTransform.convertToCartesianImage (image) |
Convert polar image to cartesian image. |
polarTransform.convertToPolarImage (image[, …]) |
Convert cartesian image to polar image. |
polarTransform.getCartesianPointsImage (…) |
Convert list of polar points from image to cartesian image points based on transform metadata |
polarTransform.getPolarPointsImage (points, …) |
Convert list of cartesian points from image to polar image points based on transform metadata |
polarTransform.ImageTransform (center, …) |
Class to store settings when converting between cartesian and polar domain |
polarTransform Module¶
-
polarTransform.
convertToCartesianImage
(image, center=None, initialRadius=None, finalRadius=None, initialAngle=None, finalAngle=None, imageSize=None, order=3, border='constant', borderVal=0.0, useMultiThreading=False, settings=None)[source]¶ Convert polar image to cartesian image.
Using a polar image, this function creates a cartesian image. This function is versatile because it can automatically calculate an appropriate cartesian image size and center given the polar image. In addition, parameters for converting to the polar domain are necessary for the conversion back to the cartesian domain.
Parameters: - image : (N, M) or (N, M, P)
numpy.ndarray
Polar image to convert to cartesian domain
Note
For a 3D array, polar transformation is applied separately across each 2D slice
Note
If an alpha band (4th channel of image is present), then it will be converted. Typically, this is unwanted, so the recommended solution is to transform the first 3 channels and set the 4th channel to fully on.
- center :
str
or (2,)list
,tuple
ornumpy.ndarray
ofint
, optional Specifies the center in the cartesian image to use as the origin in polar domain. The center in the cartesian domain will be (0, 0) in the polar domain.
If center is not set, then it will default to
middle-middle
. If the image size isNone
, the center is calculated after the image size is determined.For relative positioning within the image, center can be one of the string values in the table below. The quadrant column contains the visible quadrants for the given center. initialAngle and finalAngle must contain at least one of the quadrants, otherwise an error will be thrown because the resulting cartesian image is blank. An example cartesian image is given below with annotations to what the center will be given a center string.
¶ Value Quadrant Location in image top-left IV 1 top-middle III, IV 2 top-right III 3 middle-left I, IV 4 middle-middle I, II, III, IV 5 middle-right II, III 6 bottom-left I 7 bottom-middle I, II 8 bottom-right II 9 - initialRadius :
int
, optional Starting radius in pixels from the center of the cartesian image in the polar image
The polar image begins at this radius, i.e. the first row of the polar image corresponds to this starting radius.
If initialRadius is not set, then it will default to
0
.- finalRadius :
int
, optional Final radius in pixels from the center of the cartesian image in the polar image
The polar image ends at this radius, i.e. the last row of the polar image corresponds to this ending radius.
Note
The polar image does not include this radius. It includes all radii starting from initial to final radii excluding the final radius. Rather, it will stop one step size before the final radius. Assuming the radial resolution (see
radiusSize
) is small enough, this should not matter.If finalRadius is not set, then it will default to the maximum radius which is the size of the radial (1st) dimension of the polar image.
- initialAngle :
float
, optional Starting angle in radians in the polar image
The polar image begins at this angle, i.e. the first column of the polar image corresponds to this starting angle.
Radian angle is with respect to the x-axis and rotates counter-clockwise. The angle should be in the range of 0 to \(2\pi\).
If initialAngle is not set, then it will default to
0.0
.- finalAngle :
float
, optional Final angle in radians in the polar image
The polar image ends at this angle, i.e. the last column of the polar image corresponds to this ending angle.
Note
The polar image does not include this angle. It includes all angles starting from initial to final angle excluding the final angle. Rather, it stops one step size before the final angle. Assuming the angular resolution (see
angleSize
) is small enough, this should not matter.Radian angle is with respect to the x-axis and rotates counter-clockwise. The angle should be in the range of 0 to \(2\pi\).
If finalAngle is not set, then it will default to \(2\pi\).
- imageSize : (2,)
list
,tuple
ornumpy.ndarray
ofint
, optional Desired size of cartesian image where 1st dimension is number of rows and 2nd dimension is number of columns
If imageSize is not set, then it defaults to the size required to fit the entire polar image on a cartesian image.
- order :
int
(0-5), optional The order of the spline interpolation, default is 3. The order has to be in the range 0-5.
The following orders have special names:
- 0 - nearest neighbor
- 1 - bilinear
- 3 - bicubic
- border : {‘constant’, ‘nearest’, ‘wrap’, ‘reflect’}, optional
Polar points outside the cartesian image boundaries are filled according to the given mode.
Default is ‘constant’
The following table describes the mode and expected output when seeking past the boundaries. The input column is the 1D input array whilst the extended columns on either side of the input array correspond to the expected values for the given mode if one extends past the boundaries.
¶ Mode Ext. Input Ext. mirror 4 3 2 1 2 3 4 5 6 7 8 7 6 5 reflect 3 2 1 1 2 3 4 5 6 7 8 8 7 6 nearest 1 1 1 1 2 3 4 5 6 7 8 8 8 8 constant 0 0 0 1 2 3 4 5 6 7 8 0 0 0 wrap 6 7 8 1 2 3 4 5 6 7 8 1 2 3 Refer to
scipy.ndimage.map_coordinates()
for more details on this argument.- borderVal : same datatype as
image
, optional Value used for polar points outside the cartesian image boundaries if
border
= ‘constant’.Default is 0.0
- useMultiThreading :
bool
, optional Whether to use multithreading when applying transformation for 3D images. This considerably speeds up the execution time for large images but adds overhead for smaller 3D images.
Default is
False
- settings :
ImageTransform
, optional Contains metadata for conversion between polar and cartesian image.
Settings contains many of the arguments in
convertToPolarImage()
andconvertToCartesianImage()
and provides an easy way of passing these parameters along without having to specify them all again.Warning
Cleaner and more succint to use
ImageTransform.convertToCartesianImage()
If settings is not specified, then the other arguments are used in this function and the defaults will be calculated if necessary. If settings is given, then the values from settings will be used.
Returns: - cartesianImage : (N, M) or (N, M, P)
numpy.ndarray
Cartesian image (3D cartesian image if 3D input image is given)
- settings :
ImageTransform
Contains metadata for conversion between polar and cartesian image.
Settings contains many of the arguments in
convertToPolarImage()
andconvertToCartesianImage()
and provides an easy way of passing these parameters along without having to specify them all again.
- image : (N, M) or (N, M, P)
-
polarTransform.
convertToPolarImage
(image, center=None, initialRadius=None, finalRadius=None, initialAngle=None, finalAngle=None, radiusSize=None, angleSize=None, order=3, border='constant', borderVal=0.0, useMultiThreading=False, settings=None)[source]¶ Convert cartesian image to polar image.
Using a cartesian image, this function creates a polar domain image where the first dimension is radius and second dimension is the angle. This function is versatile because it allows different starting and stopping radii and angles to extract the polar region you are interested in.
Note
Traditionally images are loaded such that the origin is in the upper-left hand corner. In these cases the
initialAngle
andfinalAngle
will rotate clockwise from the x-axis. For simplicitly, it is recommended to flip the image along first dimension before passing to this function.Parameters: - image : (N, M) or (N, M, P)
numpy.ndarray
Cartesian image to convert to polar domain
Note
For a 3D array, polar transformation is applied separately across each 2D slice
Note
If an alpha band (4th channel of image is present), then it will be converted. Typically, this is unwanted, so the recommended solution is to transform the first 3 channels and set the 4th channel to fully on.
- center : (2,)
list
,tuple
ornumpy.ndarray
ofint
, optional Specifies the center in the cartesian image to use as the origin in polar domain. The center in the cartesian domain will be (0, 0) in the polar domain.
The center is structured as (x, y) where the first item is the x-coordinate and second item is the y-coordinate.
If center is not set, then it will default to
round(image.shape[::-1] / 2)
.- initialRadius :
int
, optional Starting radius in pixels from the center of the cartesian image that will appear in the polar image
The polar image will begin at this radius, i.e. the first row of the polar image will correspond to this starting radius.
If initialRadius is not set, then it will default to
0
.- finalRadius :
int
, optional Final radius in pixels from the center of the cartesian image that will appear in the polar image
The polar image will end at this radius, i.e. the last row of the polar image will correspond to this ending radius.
Note
The polar image will not include this radius. It will include all radii starting from initial to final radii excluding the final radius. Rather, it will stop one step size before the final radius. Assuming the radial resolution (see
radiusSize
) is small enough, this should not matter.If finalRadius is not set, then it will default to the maximum radius of the cartesian image. Using the furthest corner from the center, the finalRadius can be calculated as:
\[finalRadius = \sqrt{((X_{max} - X_{center})^2 + (Y_{max} - Y_{center})^2)}\]- initialAngle :
float
, optional Starting angle in radians that will appear in the polar image
The polar image will begin at this angle, i.e. the first column of the polar image will correspond to this starting angle.
Radian angle is with respect to the x-axis and rotates counter-clockwise. The angle should be in the range of 0 to \(2\pi\).
If initialAngle is not set, then it will default to
0.0
.- finalAngle :
float
, optional Final angle in radians that will appear in the polar image
The polar image will end at this angle, i.e. the last column of the polar image will correspond to this ending angle.
Note
The polar image will not include this angle. It will include all angle starting from initial to final angle excluding the final angle. Rather, it will stop one step size before the final angle. Assuming the angular resolution (see
angleSize
) is small enough, this should not matter.Radian angle is with respect to the x-axis and rotates counter-clockwise. The angle should be in the range of 0 to \(2\pi\).
If finalAngle is not set, then it will default to \(2\pi\).
- radiusSize :
int
, optional Size of polar image for radial (1st) dimension
This in effect determines the resolution of the radial dimension of the polar image based on the
initialRadius
andfinalRadius
. Resolution can be calculated using equation below in radial px per cartesian px:\[radialResolution = \frac{radiusSize}{finalRadius - initialRadius}\]If radiusSize is not set, then it will default to the minimum size necessary to ensure that image information is not lost in the transformation. The minimum resolution necessary can be found by finding the smallest change in radius from two connected pixels in the cartesian image. Through experimentation, there is a surprisingly close relationship between the maximum difference from width or height of the cartesian image to the
center
times two.The radiusSize is calculated based on this relationship and is proportional to the
initialRadius
andfinalRadius
given.- angleSize :
int
, optional Size of polar image for angular (2nd) dimension
This in effect determines the resolution of the angular dimension of the polar image based on the
initialAngle
andfinalAngle
. Resolution can be calculated using equation below in angular px per cartesian px:\[angularResolution = \frac{angleSize}{finalAngle - initialAngle}\]If angleSize is not set, then it will default to the minimum size necessary to ensure that image information is not lost in the transformation. The minimum resolution necessary can be found by finding the smallest change in angle from two connected pixels in the cartesian image.
For a cartesian image with either dimension greater than 500px, the angleSize is set to be two times larger than the largest dimension proportional to
initialAngle
andfinalAngle
. Otherwise, for a cartesian image with both dimensions less than 500px, the angleSize is set to be four times larger the largest dimension proportional toinitialAngle
andfinalAngle
.Note
The above logic estimates the necessary angleSize to reduce image information loss. No algorithm currently exists for determining the required angleSize.
- order :
int
(0-5), optional The order of the spline interpolation, default is 3. The order has to be in the range 0-5.
The following orders have special names:
- 0 - nearest neighbor
- 1 - bilinear
- 3 - bicubic
- border : {‘constant’, ‘nearest’, ‘wrap’, ‘reflect’}, optional
Polar points outside the cartesian image boundaries are filled according to the given mode.
Default is ‘constant’
The following table describes the mode and expected output when seeking past the boundaries. The input column is the 1D input array whilst the extended columns on either side of the input array correspond to the expected values for the given mode if one extends past the boundaries.
¶ Mode Ext. Input Ext. mirror 4 3 2 1 2 3 4 5 6 7 8 7 6 5 reflect 3 2 1 1 2 3 4 5 6 7 8 8 7 6 nearest 1 1 1 1 2 3 4 5 6 7 8 8 8 8 constant 0 0 0 1 2 3 4 5 6 7 8 0 0 0 wrap 6 7 8 1 2 3 4 5 6 7 8 1 2 3 Refer to
scipy.ndimage.map_coordinates()
for more details on this argument.- borderVal : same datatype as
image
, optional Value used for polar points outside the cartesian image boundaries if
border
= ‘constant’.Default is 0.0
- useMultiThreading :
bool
, optional Whether to use multithreading when applying transformation for 3D images. This considerably speeds up the execution time for large images but adds overhead for smaller 3D images.
Default is
False
- settings :
ImageTransform
, optional Contains metadata for conversion between polar and cartesian image.
Settings contains many of the arguments in
convertToPolarImage()
andconvertToCartesianImage()
and provides an easy way of passing these parameters along without having to specify them all again.Warning
Cleaner and more succint to use
ImageTransform.convertToPolarImage()
If settings is not specified, then the other arguments are used in this function and the defaults will be calculated if necessary. If settings is given, then the values from settings will be used.
Returns: - polarImage : (N, M) or (N, M, P)
numpy.ndarray
Polar image where first dimension is radii and second dimension is angle (3D polar image if 3D input image is given)
- settings :
ImageTransform
Contains metadata for conversion between polar and cartesian image.
Settings contains many of the arguments in
convertToPolarImage()
andconvertToCartesianImage()
and provides an easy way of passing these parameters along without having to specify them all again.
- image : (N, M) or (N, M, P)
-
class
polarTransform.
ImageTransform
(center, initialRadius, finalRadius, initialAngle, finalAngle, cartesianImageSize, polarImageSize)[source]¶ Class to store settings when converting between cartesian and polar domain
-
convertToCartesianImage
(image, order=3, border='constant', borderVal=0.0, useMultiThreading=False)[source]¶ Convert polar image to cartesian image.
Using a polar image, this function creates a cartesian image. This function is versatile because it can automatically calculate an appropiate cartesian image size and center given the polar image. In addition, parameters for converting to the polar domain are necessary for the conversion back to the cartesian domain.
Parameters: - image : (N, M) or (N, M, P)
numpy.ndarray
Polar image to convert to cartesian domain
Note
For a 3D array, polar transformation is applied separately across each 2D slice
Note
If an alpha band (4th channel of image is present), then it will be converted. Typically, this is unwanted, so the recommended solution is to transform the first 3 channels and set the 4th channel to fully on.
- order :
int
(0-5), optional The order of the spline interpolation, default is 3. The order has to be in the range 0-5.
The following orders have special names:
- 0 - nearest neighbor
- 1 - bilinear
- 3 - bicubic
- border : {‘constant’, ‘nearest’, ‘wrap’, ‘reflect’}, optional
Polar points outside the cartesian image boundaries are filled according to the given mode.
Default is ‘constant’
The following table describes the mode and expected output when seeking past the boundaries. The input column is the 1D input array whilst the extended columns on either side of the input array correspond to the expected values for the given mode if one extends past the boundaries.
¶ Mode Ext. Input Ext. mirror 4 3 2 1 2 3 4 5 6 7 8 7 6 5 reflect 3 2 1 1 2 3 4 5 6 7 8 8 7 6 nearest 1 1 1 1 2 3 4 5 6 7 8 8 8 8 constant 0 0 0 1 2 3 4 5 6 7 8 0 0 0 wrap 6 7 8 1 2 3 4 5 6 7 8 1 2 3 Refer to
scipy.ndimage.map_coordinates()
for more details on this argument.- borderVal : same datatype as
image
, optional Value used for polar points outside the cartesian image boundaries if
border
= ‘constant’.Default is 0.0
- useMultiThreading :
bool
, optional Whether to use multithreading when applying transformation for 3D images. This considerably speeds up the execution time for large images but adds overhead for smaller 3D images.
Default is
False
Returns: - cartesianImage : (N, M) or (N, M, P)
numpy.ndarray
Cartesian image (3D cartesian image if 3D input image is given)
See also
- image : (N, M) or (N, M, P)
-
convertToPolarImage
(image, order=3, border='constant', borderVal=0.0, useMultiThreading=False)[source]¶ Convert cartesian image to polar image.
Using a cartesian image, this function creates a polar domain image where the first dimension is radius and second dimension is the angle. This function is versatile because it allows different starting and stopping radii and angles to extract the polar region you are interested in.
Note
Traditionally images are loaded such that the origin is in the upper-left hand corner. In these cases the
initialAngle
andfinalAngle
will rotate clockwise from the x-axis. For simplicitly, it is recommended to flip the image along first dimension before passing to this function.Parameters: - image : (N, M) or (N, M, P)
numpy.ndarray
Cartesian image to convert to polar domain
Note
For a 3D array, polar transformation is applied separately across each 2D slice
Note
If an alpha band (4th channel of image is present), then it will be converted. Typically, this is unwanted, so the recommended solution is to transform the first 3 channels and set the 4th channel to fully on.
- order :
int
(0-5), optional The order of the spline interpolation, default is 3. The order has to be in the range 0-5.
The following orders have special names:
- 0 - nearest neighbor
- 1 - bilinear
- 3 - bicubic
- border : {‘constant’, ‘nearest’, ‘wrap’, ‘reflect’}, optional
Polar points outside the cartesian image boundaries are filled according to the given mode.
Default is ‘constant’
The following table describes the mode and expected output when seeking past the boundaries. The input column is the 1D input array whilst the extended columns on either side of the input array correspond to the expected values for the given mode if one extends past the boundaries.
¶ Mode Ext. Input Ext. mirror 4 3 2 1 2 3 4 5 6 7 8 7 6 5 reflect 3 2 1 1 2 3 4 5 6 7 8 8 7 6 nearest 1 1 1 1 2 3 4 5 6 7 8 8 8 8 constant 0 0 0 1 2 3 4 5 6 7 8 0 0 0 wrap 6 7 8 1 2 3 4 5 6 7 8 1 2 3 Refer to
scipy.ndimage.map_coordinates()
for more details on this argument.- borderVal : same datatype as
image
, optional Value used for polar points outside the cartesian image boundaries if
border
= ‘constant’.Default is 0.0
Returns: - polarImage : (N, M) or (N, M, P)
numpy.ndarray
Polar image where first dimension is radii and second dimension is angle (3D polar image if 3D input image is given)
- image : (N, M) or (N, M, P)
-
getCartesianPointsImage
(points)[source]¶ Convert list of polar points from image to cartesian image points based on transform metadata
Note
This does not convert from polar to cartesian points, but rather converts pixels from polar image to pixels from cartesian image using
ImageTransform
.The returned points are not rounded to the nearest point. User must do that by hand if desired.
Parameters: - points : (N, 2) or (2,)
numpy.ndarray
List of polar points to convert to cartesian domain
First column is r and second column is theta
Returns: - cartesianPoints : (N, 2) or (2,)
numpy.ndarray
Corresponding cartesian points from polar
points
usingImageTransform
See also
getCartesianPointsImage()
,getCartesianPoints()
,getCartesianPoints2()
- points : (N, 2) or (2,)
-
getPolarPointsImage
(points)[source]¶ Convert list of cartesian points from image to polar image points based on transform metadata
Note
This does not convert from cartesian to polar points, but rather converts pixels from cartesian image to pixels from polar image using
ImageTransform
.The returned points are not rounded to the nearest point. User must do that by hand if desired.
Parameters: - points : (N, 2) or (2,)
numpy.ndarray
List of cartesian points to convert to polar domain
First column is x and second column is y
Returns: - polarPoints : (N, 2) or (2,)
numpy.ndarray
Corresponding polar points from cartesian
points
usingImageTransform
See also
getPolarPointsImage()
,getPolarPoints()
,getPolarPoints2()
- points : (N, 2) or (2,)
-
-
polarTransform.
getCartesianPointsImage
(points, settings)[source]¶ Convert list of polar points from image to cartesian image points based on transform metadata
Warning
Cleaner and more succinct to use
ImageTransform.getCartesianPointsImage()
Note
This does not convert from polar to cartesian points, but rather converts pixels from polar image to pixels from cartesian image using
ImageTransform
.The returned points are not rounded to the nearest point. User must do that by hand if desired.
Parameters: - points : (N, 2) or (2,)
numpy.ndarray
List of polar points to convert to cartesian domain
First column is r and second column is theta
- settings :
ImageTransform
Contains metadata for conversion from polar to cartesian domain
Settings contains many of the arguments in
convertToPolarImage()
andconvertToCartesianImage()
and provides an easy way of passing these parameters along without having to specify them all again.
Returns: - cartesianPoints : (N, 2) or (2,)
numpy.ndarray
Corresponding cartesian points from polar
points
usingsettings
See also
ImageTransform.getCartesianPointsImage()
,getCartesianPoints()
,getCartesianPoints2()
- points : (N, 2) or (2,)