I am making a simple editor where the user can click on points of an image and crop out a shape. My implementation is terribly inefficient and as I'm new to qt, I have trouble deciphering all the functions on qt's docs.
QPolygonF polygon(points);
std::map<std::string, int> map = pointsHandler.getOutsideVals();
for(int i = map["Left"]; i < map["Right"]; i++){
for(int j = map["Top"]; j < map["Bottom"]; j++){
for(int n = 0; n < points.size(); n++){
if(polygon.containsPoint(QPointF(i,j), Qt::OddEvenFill)){
image.setPixelColor(QPoint(i - xOffset, j - yOffset), Qt::transparent);
}
}
}
}
painter.drawImage(xOffset,yOffset, image);
Currently how I'm doing it is looping through a rectangle given by the outer most points of the polygon. If a point is in the polygon or not I change the pixel value to be transparent. The polygon is made from the users clicked points which I then store the outer most values in a map. When I crop out large portions, it takes far to long and was I looking for some advice to make this more efficient. Thank you.
EDIT
I am now using setClipPath mentioned by G.M. and have no performance issues, however the way I found to get the job done now seems like a waste of memory. Using setClipPath(...) the best work around I found was to make multiple Qt class objects on the stack, it works great just seems like I'm working around to much stuff. Here's the updated code.
QPolygon clipPolygon = QPolygonF(points).toPolygon();
QRegion clippedRegion(clipPolygon, Qt::OddEvenFill);
QRect translatedImageRect = image.rect().translated(QPoint(xOffset, yOffset));
QRegion unClippedRegion = QRegion(translatedImageRect).subtracted(clippedRegion);
painter.save();
painter.setClipRegion(unClippedRegion, Qt::ReplaceClip);
painter.drawImage(xOffset, yOffset, image);
painter.restore();
It works great, just feel like I'm wasting memory.
You can use QPainter to make a rectangle of your image transparent.
QImage image("/home/tim/Bilder/Example.png");
QPainter painter(&image);
painter.setCompositionMode(QPainter::CompositionMode_Source);
painter.fillRect(0, 0, 10, 10, Qt::transparent);
painter.end();
image.save("/home/tim/Bilder/changed.png", "PNG");
I have an application where I copy some raw image data into a QImage directly:
QImage* img = new QImage(desc.Width, desc.Height, QImage::Format_RGB32);
for (y = 0; y < img->height(); y++)
{
memcpy(img->scanLine(y), &rawData[y * pRes->RowPitch], pRes->RowPitch);
}
return img;
Later this QImage is drawn via a call
painter.drawPixmap();
Unfortunately drawPixmap() cannot handle a QImage directly, so it first has to be converted:
m_bgImage = new QPixmap();
m_bgImage->convertFromImage(image);
Due to timing reasons I would like to drop this additional conversion step.
Thus my question: are there any function in QPixmap that allow direct image data manipulation right as in QImage?
My idea would be to start with a QPixmap from the very beginning, copy the raw image data into the QPixmap object and then use it directly.
Thanks :-)
First of all you won't need that loop to create the QImage. You can:
QImage* img = new QImage(&rawData, desc.Width, desc.Height, pRes->RowPitch * 4, QImage::Format_RGB32);
Then you can
painter.drawImage(QPointF(0,0),*img);
If there is any specific reason to use QPixmap (like QPixmap caching) you will have no other choice than convert it to QPixmap first.
I receive raw image data from server. The server uses MS Dib() function which returns in BGR format. Now, what i want to do is to read this raw data and use glDrawPixels to draw it in Linux.
I was advised that GetClrTabAddress function in MS and alike shall be used to get me the RGB values for each index of 800 by 600 image sent to me.
I do not know how to get these values using indices. Could anyone give some tips.
void func(QByteArray)
{
window_width = 800;
window_height = 600;
size = window_width * window_height;
pixels = new float[size*3];
memcpy(pixels, bytes, bytes.size());
}
void GlWidget::paintGL()
{
//! [5]
glClear(GL_COLOR_BUFFER_BIT | GL_DEPTH_BUFFER_BIT);
glDrawPixels(window_width,window_height,GL_RGB,GL_FLOAT,pixels);
}
You can use GL_BGR in glDrawPixels, which will do the conversion for you and will probably be faster since AFAIK the GPU will do the work.
QByteArray sounds like you should be using unsigned bytes/chars instead of floats, which means GL_UNSIGNED_BYTE.
I'd assert(size*3*sizeof(float) == bytes.size());.
In this case make sure to set glPixelStorei(GL_UNPACK_ALIGNMENT, 1) if your width doesn't align to the default 4-byte boundry. With GL_BGR very pixel is 3 bytes and by default each row of your pixels is assumed to be padded to the next 4-byte boundary.
[EDIT]
OK, it looks like the image uses a palette. This means every value inthe QByteArray maps to an rgb value in another array. I'm not 100% sure where the palette is and maybe it can be computed implicitly, but you mentioned GetClrTabAddress which sounds promising.
The code will then look something like this
for(int i = 0; i < size; ++i)
{
unsigned char index = btmp[i];
//and something like..
memcpy(bytes + i * 3, GetClrTabAddress() + index * 3, 3);
//or
bytes[i*3+0] = someOtherPaletteData[index].red;
bytes[i*3+1] = someOtherPaletteData[index].green;
bytes[i*3+2] = someOtherPaletteData[index].blue;
}
I need to extract frames from a video in my Qt based application. Using ffmpeg libraries I am able to fetch frames as AVFrames which I need to convert to QImage to use in other parts of my application. This conversion needs to be efficient. So far it seems sws_scale() is the right function to use but I am not sure what source and destination pixel formats are to be specified.
Came up with the following 2-step process that first converts a decoded AVFame to another AVFrame in RGB colorspace and then to QImage. It works and is reasonably fast.
src_frame = get_decoded_frame();
AVFrame *pFrameRGB = avcodec_alloc_frame(); // intermediate pframe
if(pFrameRGB==NULL) {
;// Handle error
}
int numBytes= avpicture_get_size(PIX_FMT_RGB24,
is->video_st->codec->width, is->video_st->codec->height);
uint8_t *buffer = (uint8_t*)malloc(numBytes);
avpicture_fill((AVPicture*)pFrameRGB, buffer, PIX_FMT_RGB24,
is->video_st->codec->width, is->video_st->codec->height);
int dst_fmt = PIX_FMT_RGB24;
int dst_w = is->video_st->codec->width;
int dst_h = is->video_st->codec->height;
// TODO: cache following conversion context for speedup,
// and recalculate only on dimension changes
SwsContext *img_convert_ctx_temp;
img_convert_ctx_temp = sws_getContext(
is->video_st->codec->width, is->video_st->codec->height,
is->video_st->codec->pix_fmt,
dst_w, dst_h, (PixelFormat)dst_fmt,
SWS_BICUBIC, NULL, NULL, NULL);
QImage *myImage = new QImage(dst_w, dst_h, QImage::Format_RGB32);
sws_scale(img_convert_ctx_temp,
src_frame->data, src_frame->linesize, 0, is->video_st->codec->height,
pFrameRGB->data,
pFrameRGB->linesize);
uint8_t *src = (uint8_t *)(pFrameRGB->data[0]);
for (int y = 0; y < dst_h; y++)
{
QRgb *scanLine = (QRgb *) myImage->scanLine(y);
for (int x = 0; x < dst_w; x=x+1)
{
scanLine[x] = qRgb(src[3*x], src[3*x+1], src[3*x+2]);
}
src += pFrameRGB->linesize[0];
}
If you find a more efficient approach, let me know in the comments
I know, it's too late, but maybe someone will find it useful. From here I got the clue of doing the same conversion, which looks a bit shorter.
So I created QImage which is reused for every decoded frame:
QImage img( width, height, QImage::Format_RGB888 );
Created frameRGB:
frameRGB = av_frame_alloc();
//Allocate memory for the pixels of a picture and setup the AVPicture fields for it.
avpicture_alloc( ( AVPicture *) frameRGB, AV_PIX_FMT_RGB24, width, height);
After the the first frame is decoded I create conversion context SwsContext this way (it will be used for all the next frames):
mImgConvertCtx = sws_getContext( codecContext->width, codecContext->height, codecContext->pix_fmt, width, height, AV_PIX_FMT_RGB24, SWS_BICUBIC, NULL, NULL, NULL);
And finally for every decoded frame conversion is performed:
if( 1 == framesFinished && nullptr != imgConvertCtx )
{
//conversion frame to frameRGB
sws_scale(imgConvertCtx, frame->data, frame->linesize, 0, codecContext->height, frameRGB->data, frameRGB->linesize);
//setting QImage from frameRGB
for( int y = 0; y < height; ++y )
memcpy( img.scanLine(y), frameRGB->data[0]+y * frameRGB->linesize[0], mWidth * 3 );
}
See the link for the specifics.
A simpler approach, I think:
void takeSnapshot(AVCodecContext* dec_ctx, AVFrame* frame)
{
SwsContext* img_convert_ctx;
img_convert_ctx = sws_getContext(dec_ctx->width,
dec_ctx->height,
dec_ctx->pix_fmt,
dec_ctx->width,
dec_ctx->height,
AV_PIX_FMT_RGB24,
SWS_BICUBIC, NULL, NULL, NULL);
AVFrame* frameRGB = av_frame_alloc();
avpicture_alloc((AVPicture*)frameRGB,
AV_PIX_FMT_RGB24,
dec_ctx->width,
dec_ctx->height);
sws_scale(img_convert_ctx,
frame->data,
frame->linesize, 0,
dec_ctx->height,
frameRGB->data,
frameRGB->linesize);
QImage image(frameRGB->data[0],
dec_ctx->width,
dec_ctx->height,
frameRGB->linesize[0],
QImage::Format_RGB888);
image.save("capture.png");
}
Today, I have tested directly pass the image->bit() to swscale and finally it works, so it doesn't need to copy to memory. For example:
/* 1. Get frame and QImage to show */
struct my_frame *frame = get_frame(source);
QImage *myImage = new QImage(dst_w, dst_h, QImage::Format_RGBA8888);
/* 2. Convert and write into image buffer */
uint8_t *dst[] = {myImage->bits()};
int linesizes[4];
av_image_fill_linesizes(linesizes, AV_PIX_FMT_RGBA, frame->width);
sws_scale(myswscontext, frame->data, (const int*)frame->linesize,
0, frame->height, dst, linesizes);
I just discovered that scanLine is just seeking thru the buffer.. all you need is use AV_PIX_FMT_RGB32 for the AVFrame and QImage::FORMAT_RGB32 for the QImage.
Then after decoding just do a memcpy
memcpy(img.scanLine(0), pFrameRGB->data[0], pFrameRGB->linesize[0] * pFrameRGB->height());
I had problems with the other proposed solutions as :
They did not mention freeing either AVFrame, SwsContext or the allocated buffers, which caused massive memory leaks (I had thousands of frames to handle). These problems couldn't all be solved easily as QImage relies on the underlying data, and does not copy it. If freeing the buffer directly, the QImage points to freed data and breaks. This could be solved by using QImage's cleanupFunction to free the buffer once the image is no longer needed, but with other problems it wasn't good anyways.
In some cases one of the suggestions, of passing QImage.bits directly to sws_scale, would not work as QImage are minimum 32 bit aligned. Therefore for certain dimensions it would not match the expected width by sws_scale and output each line shifted a little bit.
A third problem is that they used deprecated AVPicture elements.
I listed the problem in another question Converting an AVFrame to QImage with conversion of pixel format and in the end found a solution using a temporary buffer, which could be copied to the QImage, and then safely freed.
So see my answer for a fully working, efficient, and with no deprecated function calls, implementation : https://stackoverflow.com/a/68212609/7360943
I've found very similiar topic: how to convert an opencv cv::Mat to qimage , but it does not solve my problem.
I have function converting cv::Mat to QImage
QImage cvMatToQImg(cv::Mat& mat)
{
cv::Mat rgb;
if(mat.channels()==1)
{
cv::cvtColor(mat,rgb,CV_GRAY2BGR);
cv::cvtColor(rgb,rgb,CV_BGR2BGRA);
QImage temp = QImage((unsigned char*)(rgb.data), rgb.cols,
rgb.rows,QImage::Format_ARGB32 );
QImage returnImage = temp.copy();
return returnImage;
}
And it's works for my but I want to make it more efficient.
First: why changing 2 cvtColor functions with:
cv::cvtColor(mat,rgb,CV_GRAY2BGRA)
fails on
QImage returnImage = temp.copy()
with segfault.
Then how to eliminate copying of QImage. When I simply return temp image, I'm getting segfault.
Any other optimalizations can be done there? It's very often used function so I want to make it as fast as possible.
Your solution to the problem is not efficient, in particular it is less efficient then the code I posted on the other question you link to.
Your problem is that you have to convert from grayscale to color, or RGBA. As soon as you need this conversation, naturally a copy of the data is needed.
My solution does the conversion between grayscale and color, as well as between cv::Mat and QImage at the same time. That's why it is the most efficient you can get.
In your solution, you first try to convert and then want to build QImage around OpenCV data directly to spare a second copy. But, the data you point to is temporary. As soon as you leave the function, the cv::Mat free's its associated memory and that's why it is not valid anymore also within the QImage. You could manually increase the reference counter of the cv::Mat beforehand, but that opens the door for a memory leak afterwards.
In the end, you attempt a dirty solution to a problem better solved in a clean fashion.
It may be easiest to roll your own solution. Below is the current OpenCV implementation for going from gray to RGBA format:
template<typename _Tp>
struct Gray2RGB
{
typedef _Tp channel_type;
Gray2RGB(int _dstcn) : dstcn(_dstcn) {}
void operator()(const _Tp* src, _Tp* dst, int n) const
{
if( dstcn == 3 )
for( int i = 0; i < n; i++, dst += 3 )
{
dst[0] = dst[1] = dst[2] = src[i];
}
else
{
_Tp alpha = ColorChannel<_Tp>::max();
for( int i = 0; i < n; i++, dst += 4 )
{
dst[0] = dst[1] = dst[2] = src[i];
dst[3] = alpha;
}
}
}
int dstcn;
};
Here is where the actual cvtColor call occurs:
case CV_GRAY2BGR: case CV_GRAY2BGRA:
if( dcn <= 0 ) dcn = 3;
CV_Assert( scn == 1 && (dcn == 3 || dcn == 4));
_dst.create(sz, CV_MAKETYPE(depth, dcn));
dst = _dst.getMat();
if( depth == CV_8U )
CvtColorLoop(src, dst, Gray2RGB<uchar>(dcn));
This code is contained in the color.cpp file in the imgproc library.
As you can see, since you are not setting the dstCn parameter in your cvtColor calls, it defaults to dcn = 3. To go straight from gray to BGRA, set dstCn to 4. Since OpenCV's default color order is BGR, you'll still need to swap the color channels for it to look right (assuming you get your image data from an OpenCV function). So, it may be worth it to implement your own converter possibly following the above example, or using ypnos answer here.
Also, have a look at my other answer involving how to integrate OpenCV with Qt.
The problem is that both the cv::Mat and QImage data isn't necessarily contiguous.
New data rows in opencv start on a 32bit boundary (not sure about QImage - I think it's system dependant) so you can't copy a memeory block unless your rows happen to be exact multiples of 4bytes
See How to output this 24 bit image in Qt