Full sheet vs. details: We decided to always take images of entire herbarium sheets, and we required an image resolution sufficiently high to show details of the plants and label texts. The threshold was put to be roughly equivalent to viewing the original sheet with the naked eye.
Textual data and verification: Much of our discussion centered upon the extent and quality of the textual information accompanying the image. Clearly, the ideal would be to provide a full transcription of the original label’s text as well as any annotations found on the sheet itself, accompanied by an up-to-date interpretation of, e.g., the collection locality. As a default, we thus decided to largely publish the information as found in the herbarium, and confine textual data input to the very minimum. Currently, this comprises (i) the barcode of the specimen, (ii) the scientific name under which the specimen is stored in the herbarium, (iii) the modern country of origin (where it can be easily deduced), and (iv) the original designation of the collection site. Optionally, (v) collector and (vi) collection number are also registered.
Camera and lighting: In November 2008 we started image production with a Pentax K20D photographic camera. The major initial difficulty concerned the lighting equipment, the original lamps were of the wrong type and caused strong color deviations, furthermore they were generating too much heat which caused the CCD chip to change its properties. The equipment was replaced accordingly and we now use for all cameras two Kaiservision lamps each with 2 or 4 36W light bulbs. Selecting the appropriate lighting equipment with respect to spectrum and effect on ambient temperature is an essential prerequisite for high-quality images. The lessen we learned here was that the lens has a profound effect on the quality of the digital image, which seems to be much more pronounced than in traditional photography.
Digital Camera: Pentax K20D
Max. resolution: 16.2 Mpx
Lens Software Interface: Pentax Remote Assistant 3
Medium file size: 10MB
Image Processing Software: Adobe Photoshop Creative Suite 2
Once more, lenses had to be changed to obtain reasonable depth of field and properly focused images (especially in marginal areas of the image). For one of the cameras, stronger lights were installed to be able to use a narrower aperture for increased depth of field (thick objects). Another problem consisted in variations in the quality of the pictures between cameras as well as within the images taken by the same camera. Tests were run to explain this in terms of ambient factors, hardware used or the software setup, to no avail. Several upgrades of the software had to be installed before the system was functioning properly for any prolonged period of time. Lack of appropriate software documentation led to a delay in finding the appropriate parameters (e.g. grid factor) to adjust the software and adopt it to the chosen resolution. All this did cost substantial time for testing and implementation, but finally we obtained a working system constantly producing images of high quality (although some variation can still be observed). The resolution of specimens images is equivalent to slightly more than 600 dpi.
Copyright: We retain copyright for the images. The user is not authorized to pass on images any third party without our written consent and due mention of the copyright restrictions, and the images may not be used for commercial or non-scientific purposes. Reproduction in scientific publications is authorized on the condition that the source and the copyright are fully acknowledged.
The imaging practice
Basic setup and process: Images are taken with a constant distance and standard parameters for a given sheet size. A pre-photo is used to control focus, depth of field and color adjustments, followed by the main photo of the specimen into Photoshop. Scanning the specimen’s barcode with a barcode reader serves to name the image file and connect it to the database.
Setup of camera and lighting: A number of preparatory steps are necessary to obtain high image quality. This starts with ensuring that the camera is in a horizontal position (use a spirit level). Lighting should be maximized by adjusting distance and angle of lamps, but excessive flare must be avoided– areas with tonal values >250 in the RGB-histogram should be rare. A uniform illumination should be achieved, testing is done by scanning a white sheet and checking the white-level with a reflection densitometer, the values should vary only moderately. It is recommended to use a luxmeter to measure the illumination and to optimize accordingly. All this may be done with different lamps and arrangements of lamps to achieve an optimal setup. Since a scanning process is taking place, obviously any commotion or vibration of the object or camera must be avoided. In setting up the workplace the properties of the building must thus be considered, too. Even in a relatively solid place like our museum building, event such as slamming the door result in multicolor fault lines on the image. Places with heavy traffic or other causes of vibration may be completely unsuited for setting up the imaging equipment. Reflective objects should be positioned as close to the center as possible to avoid flare. In our case positioning the ruler towards the middle of the long margin of the sheet was sufficient, since lengh to width ratio of the herbarium sheets exceeds that of the image taken by the camera.
Focusing: Finding the optimal focus (first manually and subsequently by software) should commence with maximum aperture using a test sheet with lines. To test for depth-of-field, a 45° wedge with a millimeter scale on the slope has proven most advantageous. With thicker objects (branches etc.), the focus should be fixed at 50% of the object’s thickness, which of course should not exceed total depth of field. Using ad-hoc optimization is possible by focusing on different levels until optimal focus for the paper and the highest point in the object is reached, but this is a time-consuming process. Using a different aperture changes depth-of-field and makes re-adjustment necessary. From time to time, a check should be made if the entire area of the scan is in focus. A sheet with small print covering the entire area is useful for that purpose. It is scanned for that purpose, and checked especially in the marginal areas. Faulty areas may be the result of the camera being unleveled, of heterogeneous illumination, or of dirt on the lens.
Imaging: Adjustments made to calibrate the photographic software and the image editing program vary according to the software programs used and we are therefore not going into detail here. In general, where both programs offer adjustments, these should be set to the same values (which is sometimes rather difficult to figure out, if terminology varies). Different cameras also offer widely varying options for adjustments. Automatic settings for unsharp masking and adjusting color balance should be avoided, since our objects are not the average photographer’s choice. Using a histogram to analyse the picture has proven essential, because most image corrections can be diagnosed by looking at the histogram. For example, if the histogram is too narrow, changing the aperture and/or exposure time (scanning speed) can be used to achieve a much improved result. Correction of color saturation is achieved by adjusting the color channels in the histogram. Adjusting the white balance is often necessary because ambient light conditions influence the color quality. Color cast correction is achieved by choosing a constant white area (e.g. the white in the color scale, or part of the barcode label) and adjusting the red-green-blue values to read close to 250 in the densitometer. The scanned picture can be further optimized by changing brightness and contrast, preferably using a gradation curve. Many tutorials for improvements of digital images with different programs are available on the Internet, we recommend to set aside some time for training and studying for the personnel involved in the digitization process.
The technology is available to produce high resolution images and to make them available in a usable form, although technical improvements to speed-up the work process are needed. We are convinced that the availability of digital images of herbarium specimens (as well as other images of organisms and parts of organisms) will soon form an indispensable part of primary data provision for basic research, and that it will strongly influence the work practice of the individual taxonomist. Taxonomy today is being squeezed between growing demand and urgency for classification and identification of organisms on the one hand, and diminishing personnel resources (as measured in time available for research) on the other hand. Bringing the data to the specialist is thus a top priority for infrastructural improvements. While the physical specimen cannot be replaced as a permanent source of primary data, the features inherent in the specimen’s superficial physical appearance can be mobilized and thus shared very efficiently by using digital images on the Internet. Digital imaging of specimens is thus a major contribution to the new suprainstitutional collaborative infrastructure taxonomic research needs to tackle the challenges ahead.or contrast, colour and brightness.