Gigapixel Imaging

An overall view of a gigapixel image of a daguerreotype. Click to view the actual Gigapixel image in a new window.

Simply stated, a gigapixel image is comprised of one billion pixels or more. While modern digital SLR cameras generally use a sensor with around 20 million pixels (20 megapixels), there are only a couple of cameras to date (as of 2012) that are capable of directly imaging a gigapixel image. Most current techniques involve precisely combining high-resolution digital images as a mosaic. Due to the large file size of these ultra high-resolution images, software and browsers have been created to view and interact with gigapixel images allowing users to zoom and pan viewing the amazing detail at will.

Gigapan is a collaborative project between Carnegie Mellon University and NASA Ames Intelligent Systems Division’s Robotics Division with help from Google. Gigapan includes hardware and software options for creating gigapixel images using point and shoot cameras and DSLRs. This commercially available system uses a point-and-shoot or DSLR camera mounted on a robotic panorama head. This automates the image acquisition process, taking detailed images in an overlapping grid pattern. The package includes a stitching software and a web-based viewer. The camera moves from a fixed position, and can record a nearly spherical scene.

Microsoft Research’s Interactive Media Group created HDView for the display and interaction of ultra high-resolution images. Paired with Microsoft’s Image Composite Editor (ICE), high-resolution digital images can be stitched together and viewed at no cost in all major browsers on the Windows platform and it also performs fairly well on most browsers with Macs. HDView for the pc has an interesting feature set that functions as a high-dynamic range viewer. Several settings can accommodate the very wide dynamic range found in daguerreotypes. We have found that Microsoft’s ICE software handles image stitching quickly without the creation of artifacts and seams and handles even more challenging mosaics including IR image mosaics.

Gigapixel imaging is finding its niche with cultural heritage for better studying objects in close detail, and also in the creation of virtual museums, such as the Google Art Project . Companies like Gixcel and Gigamacro are producing robots and other machines to automate the image acquisition and image processing steps.

A single image from the 112 of a daguerreotype from a micro gigapixel image.

Gigapixel imaging can be accomplished in a range of scale from micro and macro to large scale, such as landscapes or site work. At the micro level, images are acquired with a camera and microscope and the depth of detail is dependent on the objective magnification and resolution, and number of images taken. The shallow depth of field of optical microscopy means that very flat objects are the best candidates. However, extended depth of field imaging, or z-axis scanning, can be combined with mosaic composites to create gigapixel images. The simplest set up would be to use a stereomicroscope with camera tube and some means to control object movement. However, the optics could be just as easily moved for imaging larger objects. At MCI we use a photomicroscope (Leica Z16Apo) with a variety of lighting options. The even lighting of a fiber optic or LED ring light is very useful for this work. We modified a macro focusing rail to control a stage that supported the daguerreotype. An image overlap of about 20% is generally successful. Some microscope camera manufacturers include stitching software for combining the images, but we have found Microsoft ICE’s software to be the most reliable. The “micro gigapixel” image of a daguerreotype includes 112 images that results in an exceptionally detailed image that allows viewers to zoom into dust particles, biological growth and even brush strokes found in the hand-painted details of the table cloth. The final image was 15,638 pixels x 17,750 pixels, 278 Mb tiff image.

A video walking through the viewing of the gigapixel image of a daguerreotype using HD View.

Mosaic Imaging

A single image from the macro mosaic imaging of a daguerreotype. Click to view macro mosaic image in a new window.

The merging of a number of overlapping images allows us to create a composite image of very high pixel count, which is not technically a gigapixel image. In this way we can overcome the limitation of the resolving power of the typical DSLR camera. The resulting image may only be a few times larger than the single image, but important details are now recorded. We used this approach to quickly process a higher resolution image of the daguerreotype (70x80 mm) using a macro lens.

At this level, images for mosaics are acquired with a DSLR and macro lens at close to a 1:1 ratio. Either the object or the camera needs to move in a systematic pattern that maps the entire object’s surface. It is essential to keep a fixed working distance parallel to the object surface to eliminate any planar distortion. When working with the daguerreotype, the camera was mounted on a studio stand and remained stationary, while the foam core platform supporting the object was moved. We used Canon LiveView with the camera tethered to a laptop to ensure appropriate overlap (20% is sufficient) and methodical linear movement of the object. A continuous light source was used to evenly light the entire surface, similar to the preliminary visible light with normal illumination setup. The light sources should be a fixed position in relation to the camera. If the lighting changes during the “mapping” process, the images will not stitch well, leading to seams, distortion, and other artifacts. A total of 13 tiled images was acquired and stitched together to create a macro mosaic. Adobe Photoshop was used to merge the images, but several software solutions for panorama photography can be found. The final image was 10,919 pixels x 12632 pixels, 138 Mb tiff image.

At a larger scale, a DSLR is paired with a zoom or fixed focal length lens appropriate to the project scale. This may or may not result in a gigapixel image, but it will result in a higher resolution image with finer detail.

No single technique is the best or only way to digitize objects, but instead each technique has its advantages and limitations. It is essential to understand the advantages and limitations of available imaging techniques along with clearly understanding one’s goals for digitization. Gigapixel imaging and mosaics can provide incredible detail that can be accessible for reference and research. The tiling of high-resolution digital images, whether a gigapixel image or just a mosaic, can be a great condition documentation technique for comparison, assessment, reporting for reference and conservation purposes. When looking at a daguerreotype, we can see fine details including scratches, chemical deposits, pigments, corrosion, and biological growth. Gigapixel imaging at the micro level can provide a fantastic example of condition assessment with very fine detail that can be used for before and after comparison especially when working so intensively on a single daguerreotype and not fully understanding the effects of all of the imaging and analysis techniques. The gigapixel image of the daguerreotype also provides a detailed map of the surface, which is a useful reference for analysis at the nano scale.