OpenCV image dimensions without reading entire image

Question

I'm using OpenCV and am reading gigabytes of images -- too much to fit into memory at a single time. However, I need to initialize some basic structures which require the image dimensions. At the moment I'm using imread and then freeing the image right away, and this is really inefficient.

Is there a way to get the image dimensions without reading the entire file, using opencv? If not could you suggest another library (preferably lightweight, seeing as that's all it'll be used for) that can parse the headers? Ideally it would support at least as many formats as OpenCV.

Answer 1

I don't think this is possible in opencv directly.

Although it isn't specified in the docs, Java's ImageReader.getHight (and getWidth) only parse the image header, not the whole image.

Alternatively here is a reasonable looking lightweight library that definitely only checks the headers, and supports a good amount of image formats.

Finally, if you're on Linux the 'identify ' terminal command will also give you the dimensions, which you could then read in programmatically.

Answer 2

You could use boost gil:

#include <boost/gil/extension/io/jpeg_io.hpp>

int main(int argc, char *argv[])
{
    //set/get file_path
    auto dims = boost::gil::jpeg_read_dimensions(file_path);
    int width = dims.x;
    int height = dims.y;
}

You will have to link against libjpeg, by adding -ljpeg flag to the linker. You can get some more information here.