The RAW format describes gray-scale 3D (or volume) images. In a RAW image file, the Header Size specifies the length of the header of the selected image file in bytes. A typical convention for the Header Size is 2048. A RAW file is a simple binary file, in most cases, without header information. A RAW file is created when tomographic images are taken, for example. Therefore, the person who created the file (typically the person who performed the tomography) must provide the necessary import size information.

For a file containing a three-dimensional image with NX by NY by NZ voxels with format Int, the file can consist of:

• Header Size + 1 x NX x NY x NZ bytes
• Header Size + 2 x NX x NY x NZ bytes
• Header Size + 4 x NX x NY x NZ bytes

The position of the byte (or 2 bytes or 4 bytes) in the Int file determines the position of the corresponding voxel. For a single-byte image (8 bits), all gray values from 0 to 255 can be represented. Images with two bytes per voxel (16-bit images) can represent gray values from 0 to 65,535. Finally, 32-bit images can represent approximately 4 billion gray values.

If the RAW file contains floating point values, select Float from the pull-down menu for Format in the Geometry Import dialog. For images in float format the file can consist of:

• Header Size + 4 x NX x NY x NZ bytes
• Header Size + 8 x NX x NY x NZ bytes

Indicate whether the binary data in the selected file is stored in Big Endian byte order. If this box is not checked, GeoDict will assume that the image is stored in Little Endian byte order. In a Big Endian system, the most significant bytes are stored at the smallest memory address. For more details about endianness, refer to the Wikipedia page. The creator of the RAW file must provide information about how the bytes of the image are stored.

After selecting a RAW file and incorrectly specifying the parameters in the Geometry Import dialog, clicking Quick Segmentation or Image Processing will result in a warning stating that the file length does not agree with the entered size.

In the following example, a Berea sandstone sample (Andrä et al., 2013) is used to describe the image loading workflow. Here, with the values of NX=1024, NY=1024 and NZ=1024, the file length with the Header Size (50) comes up to 1,073,741,874 bytes (Header size + NX x NY x NZ = 50 + 1024 x 1024 x 1024) in case of an 8 bit image. This is incorrect, since the actual header size is 0 and, as a result, the RAW file cannot be loaded.

A Warning appears comparing the selected file size and the calculated byte size with respect to the input values for 8 bit and 16 bit integer values. The Warning suggests an input value for the Header Size, which is necessary for RAW files with header information. Typically, values up to 2048 (bytes) may be expected here. If no suitable suggestions are provided, it is likely that the volume dimensions of the scan were entered incorrectly. In the example below, we obtained the correct header size suggestion of 0 for the 8 bit image.

We recommend including file dimensions and voxel size in the name of RAW files. Accordingly, the correct values are here extracted from the file name of the Berea sandstone structure BereaSandstone_1024_1024_1024_0.74mu.raw. In other cases, the person who performed the tomography provides the correct values in an additional readable file or as a readable header in the RAW file. These values can then be extracted using a text editor, such as Notepad++.

Click Quick Segmentation to segment the image without image processing or click Image Processing to open the image in the 3D Image Processing dialog.