jit.convolve Reference

Convolve two matrices

jit.convolve

Description

Computes the correlation between two matrices (an image and a kernel). For true convolution, rotate the kernel 180º. This may be used for typical image processing tasks like blurring, sharpening, and edge detection. By default, the convolution kernel is a 3 by 3 float32 matrix.

Examples

Matrix Operator

matrix inputs:2, matrix outputs:1
Name IOProc Planelink Typelink Dimlink Plane Dim Type
in2 adapt 0 0 0 1 1 float32
out n/a 1 1 1 1 1 char long float32 float64

More about Matrix Operators

The Jitter MOP

Since the matrix is Jitter's focus, it is not surprising that the majority of Jitter objects fall in this category of Matrix Operators. Every Matrix operator has some number of matrix inputs and some number of matrix outputs. Matrix inputs are referred to by the names "in", "in2", "in3", etc., from left to right, and matrix outputs are referred to by the names "out", "out2", "out3", etc., from left to right--i.e. the names are appended by the input/output number except for the first (leftmost) input and first (leftmost) output which are simply named "in" and "out". We will refer to the input or output name names as the "I/O-name".

Matrix inputs and outputs typically each have their own matrices internally where information is kept. This is necessary because Jitter is an asynchronous framework (i.e. all the matrices don't arrive at all inputs at the same time). Various aspects of matrix inputs and outputs can be set using the command [I/O-name] combined with one of the following suffixes: "_dim" which will set the dimensions of the specified I/O matrix, "_type" which will set the type of the specified matrix, "_planecount" which will set the plane of the specified matrix, or "_name" which will set the name of the specified matrix. There is one special case which does not have an internal matrix and this is the first input "in". This is the case since this special input actually triggers the calculation of the matrix operator, so it doesn't need to be cached until a calulation takes place, unlike the other inputs. Therefore there is no mechanism to set the dim, planecount, type, or name of "in".

Matrix operators accept what we'll refer to as "matrix args"--i.e. [planecount (int)] [ type (symbol)] [dim (int list)]. if these arguments are present, the adapt attribute will be turned off, otherwise it will be turned on. If adapt mode is turned on, each time a matrix is received in the first input, there will also be the equivalent of setting the dim, planecount, and type attributes to that of the input matrix. If the other inputs and outputs are linked to these attributes, this will affect their linked attributes as well. See the "MOP" table to determine which inputs and outputs will be linked to which attributes when adapt mode is turned on. For the leftmost input this is not applicable, and hence all columns are labelled "n/a".

The jit.matrix object is a named matrix which may be used to matrix data storage and retrieval, resampling, and matrix type and planecount conversion operations.

MOP Arguments

planecount [int]

Optional

Explicitly sets the number of planes for the output and any righthand inputs. If this is absent, the Matrix Operator will typically adapt to the lefthand incoming matrix attributes, except for special case operators.

type [symbol]

Optional

Explicitly sets the type of the matrix for the output and any righthand inputs. If this is absent, the Matrix Operator will typically adapt to the lefthand incoming matrix attributes, except for special case operators.

dimensions [list]

Optional

Explicitly sets the dimensions of the matrix for the output and any righthand inputs. If this is absent, the Matrix Operator will typically adapt to the lefthand incoming matrix attributes, except for special case operators.

MOP Attributes

adapt [int]

Matrix adaptation flag (default = 0 if matrix arguments are present, otherwise 1) When the flag is set, the jit.matrix object will adapt to the incoming matrix planecount, type, and dimensions.

[in/out]_dim [32 ints]

The matrix data dimensions (default = 1 1)

[in/out]_name [symbol]

The input or output name of the matrix (default = UID)

[in/out]_planecount [int]

The number of planes in matrix input our output data. Except in special cases, this value is equal to the planecount.

[in/out]_type [symbol]

The input or output matrix data type. Except in special cases, this value is equal to type.

outputmode [int]

Output mode (default = 1 (calculate and output matrix))
0 = No output (no calculation)
1 = Calculate and output the matrix
2 = Pass input (no calculation)
3 = Pass output (no calculation)

type [int]

The matrix data type (default = char
Supported data types are char, long, float32, or float64.

MOP Messages

bang

Equivalent to the outputmatrix message.

clear

Sets all cell values in a matrix to zero.

exportattrs

Arguments

filename [symbol]
Exports an object's current attributes values in XML format. If no filename is specified, a file dialog will open to let you choose a file.

getattributes

Sends a sequence of lists out the object's right outlet describing the object's attributes, one line per attribute. Each line listing takes the form attribute attribute-name get get-value set foo set-value data-type(s) number-of-values .

getstate

Sends a sequence of lists describing the object's state out the object's right outlet, one line per attribute. Each line listing line takes the form attribute-name attribute-value1 attribute-value2 ... attribute-valueN .

importattrs

Arguments

filename [symbol]
Imports attributes specified in XML format to set the object's attributes. If no filename is specified, a file dialog will open to let you choose a file.

jit_matrix

Arguments

matrix-name [symbol]
Handles input for the named matrix. If this messages is received in the left inlet, output is typically triggered. If this message is received in any other inlet, the data is typically cached until the jit_matrix message is received in the left inlet.

outputmatrix

Sends the matrix out the left outlet.

summary

Sends a sequence of lists describing the object and it attributes and messages out the rightmost outlet of the object. The first output line takes the form summary objectname object-name . The second and third lines describe the number of inlets and outlets for the object in the form summary (matrixinputcount/matrixoutletcount) number-of-(inlets/outlets) . The fourth line describes the matrixoutput in the form summary matrixoutput descriptor planelink planelink-value typelink typelink-value dimlink dimlink-value types data-type(s) . Each attribute for the object is then listed, one attribute per line. Each line listing takes the form summary attribute attribute-name attribute-value1 attribute-value2 ... attribute-valueN .

Attributes

boundmode [int]

Boundary case handling mode (default = 0 (clip))
0 = clip: cells which lie outside the domain of the input data will clip to the nearest cell within the domain of the input data (e.g. a convolution which would need to use cell (0,-1) would instead use cell (0,0)).
1 = wrap: cells which lie outside the domain of the input data will wrap via modulus arithmetic within the domain of the input data (e.g. a convolution on a 320 by 240 matrix which would need to use cell (0,-1) would instead use cell (0,239)).
2 = ignore: cells which lie outside the domain of the input data will not be used in calculating the convolution.

origin [2 ints]

The origin of the convolution kernel (default = 1 1)

Common Box Attributes

annotation [symbol]

Sets the text that will be displayed in the Clue window when the user moves the mouse over the object.

background [int] (default: 0)

Adds or removes the object from the patcher's background layer. background 1 adds the object to the background layer, background 0 removes it. Objects in the background layer are shown behind all objects in the default foreground layer.

color [4 floats]

Sets the color for the object box outline.

fontface [int]

Sets the type style used by the object. The options are:

plain
bold
italic
bold italic Possible values:

0 = 'regular'
1 = 'bold'
2 = 'italic'
3 = 'bold italic'

fontname [symbol]

Sets the object's font.

fontsize [float]

Sets the object's font size (in points). Possible values:

'8'
'9'
'10'
'11'
'12'
'13'
'14'
'16'
'18'
'20'
'24'
'30'
'36'
'48'
'64'
'72'

hidden [int] (default: 0)

Toggles whether an object is hidden when the patcher is locked.

hint [symbol]

Sets the text that will be displayed in as a pop-up hint when the user moves the mouse over the object in a locked patcher.

ignoreclick [int] (default: 0)

Toggles whether an object ignores mouse clicks in a locked patcher.

jspainterfile [symbol]

JS Painter File

patching_rect [4 floats] (default: 0. 0. 100. 0.)

Sets the position and size of the object in the patcher window.

position [2 floats]

g/s(set)

Sets the object's x and y position in both patching and presentation modes (if the object belongs to its patcher's presentation), leaving its size unchanged.

presentation [int] (default: 0)

Sets whether an object belongs to the patcher's presentation.

presentation_rect [4 floats] (default: 0. 0. 0. 0.)

Sets the x and y position and width and height of the object in the patcher's presentation, leaving its patching position unchanged.

rect [4 floats]

g/s(set)

Sets the x and y position and width and height of the object in both patching and presentation modes (if the object belongs to its patcher's presentation).

size [2 floats]

g/s(set)

Sets the object's width and height in both patching and presentation modes (if the object belongs to its patcher's presentation), leaving its position unchanged.

textcolor [4 floats]

Sets the color for the object's text in RGBA format.

textjustification [int]

Sets the justification for the object's text. Possible values:

0 = 'left'
1 = 'center'
2 = 'right'

varname [symbol]

Sets the patcher's scripting name, which can be used to address the object by name in pattr, scripting messages to thispatcher, and the js object.

See Also

Name Description
Working with Video in Jitter Working with Video in Jitter
jit.fastblur Blur/sharpen using optimized algorithm
jit.op Apply binary or unary operators