Todo List

Member image_geometry::PinholeCameraModel::fromCameraInfo (const sensor_msgs::CameraInfo &msg)

Calculate and use rectified ROI

Adjust P by rectified ROI instead

Member image_geometry::PinholeCameraModel::initRectificationMaps () const

For large binning settings, can drop extra rows/cols at bottom/right boundary. Make sure we're handling that 100% correctly.

Should binned resolution, K, P be part of public API?

Use rectified ROI

Member image_geometry::PinholeCameraModel::rectifyPoint (const cv::Point2d &uv_raw) const

cv::undistortPoints requires the point data to be float, should allow double

Member image_geometry::PinholeCameraModel::rectifyRoi (const cv::Rect &roi_raw) const

Actually implement "best fit" as described by REP 104.

Member image_geometry::PinholeCameraModel::unrectifyImage (const cv::Mat &rectified, cv::Mat &raw, int interpolation=cv::INTER_LINEAR) const

Implement unrectifyImage()

Member image_geometry::PinholeCameraModel::unrectifyRoi (const cv::Rect &roi_rect) const

Actually implement "best fit" as described by REP 104.

Member image_geometry::StereoCameraModel::baseline () const

Currently assuming horizontal baseline

Class PinholeTest

Tests with simple values (R = identity, D = 0, P = K or simple scaling)

Test projection functions for right stereo values, P(:,3) != 0

Tests for [un]rectifyImage

Tests using ROI, needs support from PinholeCameraModel

Tests for StereoCameraModel

Member PinholeTest::SetUp ()

Just load these from file

Member TEST_F (PinholeTest, rectifyIfCalibrated)

use forward distortion for a better test

Member TEST_F (PinholeTest, rectifyPoint)

Need R = identity for the principal point tests.