Modern imaging systems collect large amounts of data in the form of arrays of pixels. Each pixel contains information about a particular point in space. A conventional image is just a projection of a three-dimensional scene onto a two-dimensional plane. To describe the scene with more detail, one can add ranging information to each pixel, or have images of different planar cross-sections of a 3D space, and this makes the required data volumes very large. At the same time, a lot of images are ‘sparse’, i.e. large amounts of neighbouring pixels in the image carry identical information. This allows compression of the images, where the compressed image can take just a small fraction of the data volume used by an original image with minimal loss of information. This is roughly how, for example, jpeg compression works. The only drawback is that you still need to collect the complete image, and then post-process it for compression. This is not a big problem for optical cameras, which nowadays easily incorporate millions of individual detectors, and therefore can collect full images in parallel. In other frequency ranges, such arrays of detectors are a big challenge, and the images have to be collected in series. This makes image acquisition a time consuming problem, so alternative imaging techniques are therefore highly desirable.
The solution of the problem is suggested as a compressive imaging technique which in fact collects compressed information. To achieve this, the image is transmitted through a set of spatial filters, and then the signal is collected by a single or just a few detectors. The choice of the set of filters and the reconstruction of the obtained data represents the field of compressive sensing.
What authors suggest in this work is how to utilize metamaterials, or artificially designed electromagnetic structures, to create random field distributions and how to use them for the new type of imaging devices. Due to the strong frequency dispersion of the metamaterials, the field patterns created at different frequencies are vastly different, and they form the required set of ‘masks’ for compressive imaging. Therefore having just one metamaterial sample, or ‘metaimager’, and scanning different frequencies, it becomes possible to achieve compressive imaging of various schemes. The authors have also shown how such devices can be used for the imaging of 2D and 3D scenes.
You must log in to add comments.