Color digital images are rapidly becoming the medium of choice for the publication of photomicrographs in scientific literature. For example, the proportion of pages with brightfield photomicrographs published in color in Veterinary Pathology (issue 3 of each year) increased from 25% in the year 2000, to 50% in 2003, and 60% in 2006. The use of color is especially valuable for immunohistochemistry, where subtle color differences are important for image interpretation. The full application of color digital imaging is limited by several variables—two of the most challenging are color balance and even illumination.

An image with balanced color has equal illumination by all colors of the spectrum; in red, green, and blue (RGB) mode, the histogram peaks of red, green, and blue overlap, resulting in an image with a white or grey background. Modern digital cameras designed for photomicroscopy have a white balance feature that eliminates most problems with color balance, but these cameras can cost more than U.S. $10,000. Digital consumer cameras adapted for photomicroscopy cost about one-tenth as much, but they might not have a white balance feature that adequately compensates for the needs of brightfield microscopy.

Achieving even illumination throughout the field is problematic with color or black and white images, even with the best brightfield microscopes. The most obvious problem—especially at low magnification—is vignetting, in which the outer edges of the image are significantly darker than the center. Some modern digital cameras designed for photomicroscopy have a shading correction feature that decreases problems with vignetting; also, image processing software can be used to partly correct vignetting of a single image (1). However, vignetting is most effectively eliminated through use of a corrective blank field (2,3). The process involves capturing a tissue image, locking the aperture and exposure time, and then capturing a second image after moving the stage to an area of the slide that contains no tissue, but is otherwise identical to the image (i.e., same lighting, coverslip, and mounting medium). The blank image captures all aberrations in color balance and illumination that are not inherent in the stained tissue. Intuitively, color and illumination aberrations in the tissue image will be eliminated by subtraction of identical aberrations in the blank image; the resultant image will include only the tissue with a uniform white background. However, the actual image processing to achieve this goal is not as intuitive, requiring division of information in the tissue image by the blank image, followed by multiplication by a constant (2,3).

Methods available for blank-field correction all have limitations that prevent their widespread use. The free program Image Arithmetic 2.5 (2) works well, but it cannot process tagged image file format (TIFF) files. Regitnig et al. (2) provided instructions for image division using the free program Scion Image (the Microsoft® Windows® version of NIH image), but their instructions did not work on Scion Image version Alpha 4.0.3.2. Also, Scion Image cannot process Joint Photographic Experts Group (JPEG) files. Another limitation of freeprograms is that it can be expensive to get them tested and approved for loading onto networked computers. Also, blank-field correction has not been widely adopted, because the process has not been described for the image-processing program that most people actually use: Adobe® Photoshop®.

Color balance and even illumination can both be achieved with blank-field correction using image division that is part of the color dodge feature of either the full version (tested versions 5.5, CS, and CS2) or the less expensive consumer version (tested Photoshop Elements 3.0) of Adobe Photoshop (Figure 1, Table 1, and Protocol 1). All tested versions of Photoshop can process a wide variety of file formats, including JPEG, bitmap, and TIFF files. The protocol works equally well with black and white images. Images in Figure 1 and Figure 2 were captured using a Nikon DS-5M camera with a DS-L1 control unit and a Nikon Plan Apo 4× objective lens, numerical aperture (NA) 0.2. Images were captured at 2560×1920 pixel resolution and saved as bitmap files (file size 14.4 MB). The original tiled image used to create Figure 2B has a pixel resolution that can be printed at a 300-dots per inch (dpi) resolution at a full poster size of 127×31 cm. In Figure 1, the color dodge feature of Adobe Photoshop changes the bright areas of the image progressively more than the dark areas, resulting in a wider range of color intensity in the final image (Table 1). By comparison, the linear dodge feature uses image addition to produce a final image with a range of color intensity that is usually less than the original image, and the resulting final image appears overexposed. Differences in these two features are most obvious for dark staining tissues (Table 1). For presentation of brightfield photomicrographs, image division (color dodge feature) results in the best final image. Image addition (linear dodge feature) might be the best choice when quantitative image analysis is used to determine the absolute difference between bright and dark areas of an image.

Figure 1. Images for achieving blank-field correction described in Protocol 1 (A–D). (Click to enlarge)

Figure 2. Blank-field correction eliminates vignetting within a 6×2 tiled image of a normal pink salmon alevin, mid-sagittal paraffin section with hematoxylin and eosin (HE) stain (Click to enlarge)

Table 1. Red, Green, and Blue (RGB) Values For Pixels From Identical Pixel Coordinates (Click to enlarge)

Blank-field correction significantly decreases artifacts in digital images using an intuitive method that is readily repeatable and scientifically authentic (4). Another advantage of blank-field correction is that it allows for seamless stitching of multiple tiled images. This process can be used to produce high-resolution large images for posters, and the greater resolving power of higher magnification objective lenses used for each tile increases the resolution of the final overview image. Several vendors offer systems for automated image tiling or slide scanning, but these expensive systems do not produce better results than manual blank-field correction and stitching.

The use of blank-field correction for digital microscopy also has at least two challenges. First, it is difficult to find a completely clean blank image on a microscope slide, particularly when using low-power objective lenses. Because artifacts on the blank image would introduce artifacts into the tissue image (i.e., they would incorrectly make dark pixels brighter), it is best to remove discrete artifacts (such as dust, human epidermal cells, etc.) from the blank image (e.g., using the Photoshop rubber stamp tool) before inverting the blank image; to document these modifications, both the original and corrected blank images can be saved. Similar discrete artifacts should not be removed from the tissue image, because their removal might alter the data contained in the tissue image. An option for capturing a clean blank image is to remove the slide from the microscope stage and photograph the blank image without a slide (2).

The second challenge of blank-field correction is that slight variations in image intensity between the blank and tissue images will mean that in the final image about half of the pixels with no tissue will be off-scale, limited to the maximum value of 255 on the RGB scale. Off-scale loss of information can be prevented by slightly brightening the blank image before it is inverted; however, this problem might not be worth correcting, because pixels that contain tissue rarely go off-scale after correction.