Main Content

Use Prebuilt MATLAB Interface to OpenCV

The Computer Vision Toolbox™ Interface for OpenCV provides a prebuilt MATLAB® interface to the OpenCV library that you can use to directly call OpenCV functions from MATLAB without writing a C /C++ code.

Note

The Computer Vision Toolbox Interface for OpenCV in MATLAB supports OpenCV version 4.2.0 and 4.5.0. The interface does not provide GPU support.

The Computer Vision Toolbox Interface for OpenCV also provides MATLAB functions to:

  • Create MATLAB objects that represent Mat, UMat, OpenCV base classes pointed by a smart pointer, and OpenCV arrays.

    MATLAB FunctionDescription
    createMatReturns Mat objects that represent the OpenCV cv::Mat data structure, and the associated input, output, or input-output array class.
    createUMatReturns UMat objects that represent the OpenCV cv::UMat data structure, and the associated input, output, or input-output array class.
    clibArrayReturns an object that represents OpenCV native arrays or std::vector types.
    getBasePtrReturns an object that represents an OpenCV base class pointed by a smart pointer.

  • Export OpenCV outputs to MATLAB supported formats.

    MATLAB FunctionDescription
    getImageReads and exports images stored in OpenCV Mat and UMat objects to a matrix or 3-D array in the MATLAB workspace.
    keyPointsToStructExports the keypoints returned by an OpenCV keypoint detector as a MATLAB structure.
    rectToBboxExports the parameters returned by an OpenCV rectangle class to a vector in the MATLAB workspace. The parameters are exported as bounding box coordinates in one-based indexing.
    underlyingValueReturns the underlying numeric values for OpenCV enumeration objects.

The OpenCV functions in the prebuilt library do not return errors except during the run-time.

Call MATLAB Functions

To call a MATLAB function in the Computer Vision Toolbox Interface for OpenCV support package, add the package name vision.opencv.util to the import list and then call the MATLAB function. For example:

import vision.opencv.util.*
[ocvMat,ocvArray] = createMat;
You can also add the partial package name vision.opencv to the import list and call the MATLAB function by prepending util to the function name. For example:
import vision.opencv.*
[ocvMat,ocvArray] = util.createMat;

Alternatively, you can also call the MATLAB function by prefixing it with the full package name. Use this syntax to import a specific function without importing every function in the package.

[ocvMat,ocvArray] = vision.opencv.util.createMat;

Call Functions in OpenCV Library

To call a function or class in the prebuilt MATLAB interface to the OpenCV library, add the library name clib.opencv to the import list. Then call an OpenCV function by prefixing the function name with a namespace.

import clib.opencv.*
retVal = namespace.funcname(arg1,arg2,…,argN)
namespace is the namespace of the function, funcname is the OpenCV function name, and arg1,arg2,...,argN represents the input arguments for the OpenCV function. retVal is the output argument.

Alternatively, you can also call an OpenCV function by adding the package name as a prefix as shown here.

retVal = clib.opencv.namespace.funcname(arg1,arg2,…,argN)

Note

The interface supports only the cv and cvflann namespaces. For example, you can call a function in the cv namespace by using this syntax:

retVal = clib.opencv.cv.funcname(arg1,arg2,…,argN)

Display Help for MATLAB Functions

To view information about the MATLAB functions in the support package, use doc or help.

Type one of these commands in MATLAB command window.

doc vision.opencv.util.

or

help vision.opencv.util.

Then press Tab. This command loads the package, and MATLAB displays the list of functions in the package.

To view information about the package in the Help browser, type this command in the MATLAB command window, and then press Enter.

doc vision.opencv.util

Display Help for MATLAB Interface to OpenCV Library

Use these MATLAB functions to view information about the classes and the functions in the Prebuilt MATLAB interface to OpenCV library.

  • doc and help — View classes and functions in the OpenCV library.

  • methods — View constructor, method, and package function names for a class.

  • methods with '-full' option — View constructor, method, and package function signatures.

  • methodsview — View a table representation of method signatures. The methodsview window enables you to find information without scrolling through the Command Window.

Display the classes and package functions by entering this command and then pressing Tab.

doc clib.opencv.

This command loads the package, and MATLAB displays a list of the package members. To view the package, press Backspace to remove the period, then press Enter. MATLAB displays all the classes and functions in the OpenCV library.

To display class methods, call the methods function for that class. For example, to see the methods of the Affine3d class, enter this command.

methods clib.opencv.cv.Affine3d
Methods for class clib.opencv.cv.Affine3d:

Affine3d     eq           gt           le           ne           rotation     translate    
concatenate  ge           inv          lt           rotate       rvec         translation  

Static methods:

Identity     

Methods of clib.opencv.cv.Affine3d inherited from handle.

To display method signatures for a class, call the methodsview function for that class. For example:

methodsview clib.opencv.cv.Affine3d

This command opens a window that displays the methods and information about their arguments and returned values.

Limitations

The prebuilt MATLAB interface to the OpenCV library does not support functionalities that contain these following data types or language features.

  • Any type with a size greater than 64 bits, such as long double

  • Arrays of character types (wchar_t, char16_t, char32_t)

  • References to a pointer, such as int*&

  • Pointers or arrays of type std::string

  • Pointers or references to enumerations

  • Reference data members

  • void* data members

  • std smart pointers other than std::shared_ptr

  • Static data members

  • ** pointers, except:

    • MATLAB supports char**

    • MATLAB supports ** pointers to custom classes used as function or ​method parameter types.

  • Multilevel pointers, such as type***

  • C function pointers and std::function as function return types or data members. You also cannot pass a MATLAB function as input to C function pointers or an std::function parameter.

  • Class templates with incomplete or no instantiations

  • User-defined data type union

  • Arrays of type std::vector

  • Types defined in the std namespace, except these supported types:

    • std::string

    • std::wstring

    • std::u16string

    • std::u32string

    • std::vector

    • std::shared_ptr

    • std::function

Related Topics