Image.h

Go to the documentation of this file.

// ----------------------------------------------------------------------------
// -                        Open3D: www.open3d.org                            -
// ----------------------------------------------------------------------------
// Copyright (c) 2018-2023 www.open3d.org
// SPDX-License-Identifier: MIT
// ----------------------------------------------------------------------------
 
#pragma once
 
#include <Eigen/Core>
#include <memory>
#include <vector>
 
#include "open3d/geometry/Geometry2D.h"
#include "open3d/utility/Logging.h"
 
namespace open3d {
 
namespace camera {
class PinholeCameraIntrinsic;
}
 
namespace geometry {
 
class Image;
 
typedef std::vector<std::shared_ptr<Image>> ImagePyramid;
 
class Image : public Geometry2D {
public:
    enum class ColorToIntensityConversionType {
        Equal,
        Weighted,
    };
 
    enum class FilterType {
        Gaussian3,
        Gaussian5,
        Gaussian7,
        Sobel3Dx,
        Sobel3Dy
    };
 
public:
    Image() : Geometry2D(Geometry::GeometryType::Image) {}
    ~Image() override {}
 
public:
    Image &Clear() override;
    bool IsEmpty() const override;
    Eigen::Vector2d GetMinBound() const override;
    Eigen::Vector2d GetMaxBound() const override;
 
    bool TestImageBoundary(double u, double v, double inner_margin = 0.0) const;
 
public:
    virtual bool HasData() const {
        return width_ > 0 && height_ > 0 &&
               data_.size() == size_t(height_ * BytesPerLine());
    }
 
    Image &Prepare(int width,
                   int height,
                   int num_of_channels,
                   int bytes_per_channel) {
        width_ = width;
        height_ = height;
        num_of_channels_ = num_of_channels;
        bytes_per_channel_ = bytes_per_channel;
        AllocateDataBuffer();
        return *this;
    }
 
    int BytesPerLine() const {
        return width_ * num_of_channels_ * bytes_per_channel_;
    }
 
    std::pair<bool, double> FloatValueAt(double u, double v) const;
 
    static std::shared_ptr<Image>
    CreateDepthToCameraDistanceMultiplierFloatImage(
            const camera::PinholeCameraIntrinsic &intrinsic);
 
    std::shared_ptr<Image> CreateFloatImage(
            Image::ColorToIntensityConversionType type =
                    Image::ColorToIntensityConversionType::Weighted) const;
 
    template <typename T>
    T *PointerAt(int u, int v) const;
 
    template <typename T>
    T *PointerAt(int u, int v, int ch) const;
 
    template <class T>
    T *PointerAs() const {
        if (sizeof(T) != bytes_per_channel_) {
            utility::LogError("sizeof(T) != byte_per_channel_: {} != {}.",
                              sizeof(T), bytes_per_channel_);
        }
        return (T *)(data_.data());
    }
 
    std::shared_ptr<Image> ConvertDepthToFloatImage(
            double depth_scale = 1000.0, double depth_trunc = 3.0) const;
 
    std::shared_ptr<Image> Transpose() const;
 
    std::shared_ptr<Image> FlipHorizontal() const;
    std::shared_ptr<Image> FlipVertical() const;
 
    std::shared_ptr<Image> Filter(Image::FilterType type) const;
 
    std::shared_ptr<Image> Filter(const std::vector<double> &dx,
                                  const std::vector<double> &dy) const;
 
    std::shared_ptr<Image> FilterHorizontal(
            const std::vector<double> &kernel) const;
 
    std::shared_ptr<Image> Downsample() const;
 
    std::shared_ptr<Image> Dilate(int half_kernel_size = 1) const;
 
    Image &LinearTransform(double scale = 1.0, double offset = 0.0);
 
    Image &ClipIntensity(double min = 0.0, double max = 1.0);
 
    template <typename T>
    std::shared_ptr<Image> CreateImageFromFloatImage() const;
 
    static ImagePyramid FilterPyramid(const ImagePyramid &input,
                                      Image::FilterType type);
 
    ImagePyramid CreatePyramid(size_t num_of_levels,
                               bool with_gaussian_filter = true) const;
 
    std::shared_ptr<Image> CreateDepthBoundaryMask(
            double depth_threshold_for_discontinuity_check = 0.1,
            int half_dilation_kernel_size_for_discontinuity_map = 3) const;
 
protected:
    void AllocateDataBuffer() {
        data_.resize(width_ * height_ * num_of_channels_ * bytes_per_channel_);
    }
 
public:
    int width_ = 0;
    int height_ = 0;
    int num_of_channels_ = 0;
    int bytes_per_channel_ = 0;
    std::vector<uint8_t> data_;
};
 
}  // namespace geometry
}  // namespace open3d