OpenCV 2.4.7 conflict with Qt 4.8.1 QImage

本文关键字：Qt QImage conflict OpenCV with 更新时间：2023-10-16

Ubuntu 12.04，Qt 4.8，OpenCv 2.4.3，PCL 1.6。通过Qt Creator 2.5.2开发所有内容。全部采用64位版本。

我直到现在才遇到这个问题，因为我通过"cv:：imread(filename)"处理磁盘中的图像文件，然后将其数据传输到QImage->QPixmap->QLabel。没问题。

现在，我需要将一些图像加载到Qt资源文件中。我通过QtCreator创建了正确的.qrc，并在其中复制和注册了图像文件。好吧，如果我试图从资源加载图像，它不起作用：

QImage qi = QImage(":images/red.jpg");
if (qi.isNull())
{
QMessageBox::warning(this,"hey","hey");
}

当然，一开始我整天都在为资源名称而挣扎。直到我决定启动另一个项目，并复制粘贴资源文件和加载该QImage的代码。在测试项目中，它工作正常。(相同的计算机，相同的IDE和环境，但没有OpenCV链接)。看到这一点，我决定尝试这段代码：

QString fn = QFileDialog::getOpenFileName(this,"Load image");
qi = QImage(fn);
if (qi.isNull())
{
QMessageBox::warning(this,"hey","hey");
}

在原始代码中，这不起作用！fn加载了正确的文件名，但"qi=QImage(fn)；"失败。在测试程序中，这与预期的一样工作。

这是一个大问题，因为我想在窗口中设置一些图标，并为几个小部件和QSplashScreen设置一些图像。将图像文件复制到生成的二进制文件文件夹中并加载它们是不太可接受的，假设用户不会对它们做任何操作。。。

我必须再次指出，这不是Qt资源文件名的问题。

对此有什么变通方法或想法吗？

PD：我试过QPixmap("：images/red.jpg")；但程序执行失败。

PD：一些项目设置：

ResFiles.qrc:

<RCC>
<qresource prefix="/">
<file>images/red.png</file>
</qresource>
</RCC>

QOP.proj文件：

#-------------------------------------------------
#
# Project created by QtCreator 2013-03-09T23:56:16
#
#-------------------------------------------------
QT += core gui
greaterThan(QT_MAJOR_VERSION, 4): QT += widgets
TARGET = QOP
TEMPLATE = app
SOURCES += main.cpp
mainwindow.cpp 
[several files more...]
HEADERS += mainwindow.h 
[several files more...]
FORMS += mainwindow.ui 
camwindow.ui 
dialogsetupcam.ui 
dialogtestcam_sr4k.ui 
dialogtestcam_pmdnano.ui 
dialogcommand_pmdnano.ui 
dialogcreatecam.ui
CONFIG += link_pkgconfig
PKGCONFIG += opencv
CONFIG += link_pkgconfig
PKGCONFIG +=
#Do not use this configuration to get PCL, some sublibraries aren't properly accesible, in example Visualizer
#CONFIG += link_pkgconfig
#PKGCONFIG += pcl_io-1.6
#PCL Headers
INCLUDEPATH += /usr/include/pcl-1.6
/usr/include/flann/
/usr/include/eigen3/
/usr/include/openni/
#PCL Libraries
LIBS += -lpcl_common
-lpcl_features
-lpcl_filters
-lpcl_geometry
-lpcl_io
-lpcl_kdtree
-lpcl_keypoints
-lpcl_octree
-lpcl_registration
-lpcl_sample_consensus
-lpcl_search
-lpcl_segmentation
-lpcl_surface
-lpcl_tracking
-lpcl_visualization
#VTK Headers
INCLUDEPATH += /usr/include/vtk-5.8
#VTK Libraries
LIBS += -lQVTK
-lvtkalglib
-lvtkCharts
-lvtkCommon
-lvtkDICOMParser
-lvtkexoIIc
-lvtkFiltering
-lvtkftgl
-lvtkGenericFiltering
-lvtkGeovis
-lvtkGraphics
-lvtkHybrid
-lvtkImaging
-lvtkInfovis
-lvtkIO
-lvtkmetaio
-lvtkParallel
-lvtkproj4
-lvtkQtChart
-lvtkRendering
-lvtksys
-lvtkverdict
-lvtkViews
-lvtkVolumeRendering
-lvtkWidgets
RESOURCES += 
ResFiles.qrc
OTHER_FILES += 
images/red.png

我想的一个解决方法是用一些QFile类加载这些图像，并将其"转移"到QImage。这能做到吗？

我用这种方式解决了这个问题：

//This WON'T WORK due to problems caused by presence of OpenCV library; it interferes with normal QImage loading from QResources
/*
QImage logoImage(":/images/Logo.png");
*/
//Read file rawly into RAM (workaround to OpenCV/QImage issue)
bool isImageInRAM = false;
QFile       file(":/images/Logo.png"); //Apparently there's no difference between ":/images/QOPLogo.png" and ":images/Logo.png"
QByteArray  rawImageFileInRAM;
if (file.open(QIODevice::ReadOnly))
{
rawImageFileInRAM = file.readAll();
file.close();
isImageInRAM = true;
}
else
checkFailed(0,QObject::tr("Error loading image from resource: %1").arg(QObject::tr("Logo"))); //checkFailed is a custom method that shows message errors
//This WON'T WORK due to problems caused by presence of OpenCV library; it interferes with normal QImage loading from memory (yes, too U_U)
/*
QImage logoImage = QImage::fromData(rawImageFileInRAM);
*/
std::vector<uchar> buffInRAM(rawImageFileInRAM.data(),rawImageFileInRAM.data()+rawImageFileInRAM.size());
cv::Mat imgMatFromRAM = cv::imdecode(cv::Mat(buffInRAM),CV_LOAD_IMAGE_COLOR);
cv::Mat cvRGBLogoImage(imgMatFromRAM.rows,imgMatFromRAM.cols,CV_8UC3);
cv::cvtColor(imgMatFromRAM,cvRGBLogoImage,CV_BGR2RGB); //Convert BGR->RGB
QImage logoImage((const unsigned char*)cvRGBLogoImage.data,cvRGBLogoImage.cols,cvRGBLogoImage.rows,cvRGBLogoImage.step,QImage::Format_RGB888);
//At this point logoImage already has image from resources ¡Yeaahhhh!!