Previous research has indicated that coded masks with open fractions <0.5 are optimal for imaging some types of far-field scenes. The open fraction, in this case, refers to the ratio of open elements in the mask, with values <0.5 considered as low open fraction. Research is limited by the sparsity of <0.5 open fractions masks; thus a further 94 lower open fraction arrays are calculated and presented. These include the dilute uniformly redundant array and singer set, along with information on imaging potential, array sizes, and open fractions. Signal-to-noise ratio reveals the 0.5 open fraction modified uniformly redundant array to be the optimal coded mask for near-field x-ray backscatter imaging, over the lower open fraction singer set, dilute uniformly redundant and random array.
Coded masks (CM) often lack a self-supporting structure that is difficult to manufacture without recourse to drilled holes in place of ideal square apertures, degrading imaging properties. An alternative approach is presented with three-dimensional (3-D) printed CM molds cast with a radio-opaque material that allows square elements to be retained. Two methods are presented; hot casting a bismuth alloy (density 8.6 g cm − 3) and cold casting with tungsten powder/epoxy resin (densities 9.6 and 10.6 g cm − 3). A critical review of 3-D printed-CM fabrication along with some typical x-ray backscatter images is presented. A signal-to-noise ratio from both the machined tungsten and cold cast 3-D printed mask were comparable, with the former having a slight advantage. Also, 3-D printed cold cast masks were found to be more economical and easier to rapid prototype over traditional drilled tungsten masks.
Many different mask patterns can be used for X-ray backscatter imaging using coded apertures, which can find application in the medical, industrial and security sectors. While some of these patterns may be considered to have a self-supporting structure, this is not the case for some of the most frequently used patterns such as uniformly redundant arrays or any pattern with a high open fraction. This makes mask construction difficult and usually requires a compromise in its design by drilling holes or adopting a no two holes touching version of the original pattern. In this study, this compromise was avoided by 3D printing a support structure that was then filled with a radiopaque material to create the completed mask. The coded masks were manufactured using two different methods, hot cast and cold cast. Hot casting involved casting a bismuth alloy at 80°C into the 3D printed acrylonitrile butadiene styrene mould which produced an absorber with density of 8.6 g cm<sup>-3</sup>. Cold casting was undertaken at room temperature, when a tungsten/epoxy composite was cast into a 3D printed polylactic acid mould. The cold cast procedure offered a greater density of around 9.6 to 10 g cm<sup>-3</sup> and consequently greater X-ray attenuation. It was also found to be much easier to manufacture and more cost effective. A critical review of the manufacturing procedure is presented along with some typical images. In both cases the 3D printing process allowed square apertures to be created avoiding their approximation by circular holes when conventional drilling is used.
Single sided radiographic imaging would find great utility for medical, aerospace and security applications. While coded apertures can be used to form such an image from backscattered X-rays they suffer from near field limitations that introduce noise. Several theoretical studies have indicated that for an extended source the images signal to noise ratio may be optimised by using a low open fraction (<0.5) mask. However, few experimental results have been published for such low open fraction patterns and details of their formulation are often unavailable or are ambiguous. In this paper we address this process for two types of low open fraction mask, the dilute URA and the Singer set array. For the dilute URA the procedure for producing multiple 2D array patterns from given 1D binary sequences (Barker codes) is explained. Their point spread functions are calculated and their imaging properties are critically reviewed. These results are then compared to those from the Singer set and experimental exposures are presented for both type of pattern; their prospects for near field imaging are discussed.