Reading binary image data

The image model

The image model used by libpninexus is rather simple. We assume that an image is a regular 2 dimensional array with nx elements along the first dimension and ny elements along the second dimension. In general the x-direction is the slow direction and the y-direction the fast one. This means typically that nx is typically associated with the number of rows or strips of an image and ny with the number of pixels per strip or row.

The elements on the images grid are pixels. Every pixel is again subdivided in a set of channels. For instance an RGB image would have a red-, green- and blue-channel. Additionally a file can store multiple images (for instance TIFF). The interface takes care about this.

The base class for all image readers is image_reader. In addition there are two additional classes providing basic image meta-data

  • image_info - with general information about an image (number of pixels along each dimension, total number of bits per pixel)

  • and image_channel_info - providing information about the individual channels stored in each image.

Image reader

Currently there are two classes implementing the image_reader interface:

As both have a similar interface only the CBF reader will be discussed here.

In order to use the CBF reader you have to include pni/cbf/cbf_reader.hpp in your source file (for the TIFF reader use pni/tiff/tiff_reader.hpp).

#include <pni/cbf/cbf_reader.hpp>
#include <pni/image_info.hpp>

pni::cbf_reader reader("image_00001.cbf");

We can check whether there are several images in the file and have a look on their basic structure with

for(size_t index=0;index<reader.nimages();index++)
   std::cout<<reader.info(index)<<std::endl;

To read data we could use

using Frame = std::vector<pni::float32>;

//read the first channel from the first image in the file
auto frame = reader.image<Frame>(0,0);