Template Class Image

Defined in File image.hpp

Inheritance Relationships

Base Type

public cv::Mat

Class Documentation

template<typename T, int l> class Image : public cv::Mat

Base class for image data used in 3D image processing.

The lib3d::types::Image class provides a more convenient and consistent use of cv::Mat. It is templated with a typename T and a number of channels l (layers). It subclasses cv::Mat.

A number of manifistations, which are key to the lib3D and 3D processing, are predefined with a fixed datatype and a fixed number of channels:

ColorImage: three-channel RGB image with pixel data of type uchar. Content string: ‘RGB’
GrayscaleImage: single-channel grayscale image with pixel data of type uchar. Content string: ‘Luminance’
DepthMap: single-channel image with pixel data of type float. Content string: ‘Depth’
DDepthMap: single-channel image with pixel data of type double. Content string: ‘Depth’
DisparityMap: single-channel image with pixel data of type float. Content string: ‘Disparity’
DDisparityMap: single-channel image with pixel data of type double. Content string: ‘Disparity’
ConfidenceMap: single-channel image with pixel data of type float. Content string: ‘Confidence’
DConfidenceMap: single-channel image with pixel data of type double. Content string: ‘Confidence’
NormalMap: three-channel image with pixel data of type float. Content string: ‘Normal’
DNormalMap: three-channel image with pixel data of type double. Content string: ‘Normal’

Note: The content string of the generic lib3d::types::Image class will be initialized with ‘Image’. This can be changed with setContentStr().

The lib3d::types::Image class can be used in the same way as cv::Mat. However, the member function valueAt() provides a ‘type-aware’ access to the image data. In this, the data can be indexed by either passing the x, y and c (channel) coordiantes separately, or by passing a cv::Point2i or cv::Point3i. Dependent on the number of channels the data can be accessed with a two-fold (i.e. x-y or cv::Point2i) or a three-fold (i.e. x-y-c or cv::Point3i) index.

Note: When passing the indices separately, the horizontal index (x) comes before the vertical index (y). This is different to using cv::Mat::at() but provides a more intuitive interface.

An instance of a cv::Mat can directly be assigned to lib3d::types::Image, by emplyoing the Image::operator=(const Mat& m) function. This is particularly needed for using functions returning a cv::Mat, e.g. cv::imread().

Note: In the Image::operator=(const Mat& m) function it is evaluated whether the type of the passed cv::Mat corresponds to the type of the Image instance. If this is not the case, the cv::Mat is not assigned to the lib3d::types::Image.

Template Parameters:

T – Type name of pixel data (e.g. uchar, float, …).
l – Number of channels (layers) inside the image.

Public Functions

inline Image(): Default constructor.

inline Image(const std::string iContentStr): Zero initialization constructor creating an empty image.

inline Image(const int &iWidth, const int &iHeight, const std::string iContentStr = "Image"): Class constructor creating a zero-initialized image of size iWidth \(\times\) iHeight.

inline Image(const cv::Size &iSize, const std::string iContentStr = "Image"): Class constructor creating a zero-initialized image of size iSize.

inline virtual ~Image(): Destructor.

inline Image &operator=(const Mat &m): Assignment operator allowing to assign an instance of cv::Mat to an instance of lib3d::types::Image.

Note

In this, it is evaluated whether the type of the passed cv::Mat corresponds to the type of the image instance. If this is not the case, the image object is kept unchanged.

template<int ch = l> inline std::enable_if<(ch == 1), T&>::type valueAt(int x, int y)

Various methods providing easy access to the pixel data. Depending on the datatype and the channel number of the image the parameters and the return value differ.

Parameters:

x – [in] X-coordinate of the pixel to access.
y – [in] Y-coordinate of the pixel to access.

Returns:

Reference of type T if the Image has 1 channel.

template<int ch = l> inline std::enable_if<(ch == 1), constT&>::type valueAt(int x, int y) const