Presented paper deals with the image processing for the new 3D voxel detector. The detector is mainly intended for particle tracking applications, but it has also advantages in conventional X-ray transmission radiography. Imaging with such device brings benefits such as higher detection efficiency, presence of directional information and mapping of beam-hardening effects. To get these additional imaging possibilities, the raw image data acquired by the voxel detector have to be processed by a method of image matching based on the Lucas-Kanade algorithm in conjunction with a cross-correlation method. The main result of the image data processing is an assembled image ready for further analysis. The secondary outcome is the information about the geometry of the imaging setup (i.e. measured position of the detector relative to the X-ray source). Results together with evaluation of the techniques are demonstrated on images of real samples.