| |
| ||||||||||||||||||||||||||||
Proactive disclosure Print version   | |  Geochemical Modeling EDS Image Processing and Enhancement
Image processing includes: 1) conversion of grayscale images to RGB color, 2) setting threshold levels, 3) saturation of the features into binary images, and 4) setting cutoff values to remove noise in the binary image. Image enhancement eliminates aberration related to surface imperfection on the epoxy puck and to the manipulation of grain edge effects. EDS-TIFF files were processed and enhanced using Adobe PhotoshopTM 5.5 (Adobe 1999) and Fovea ProTM 1.0 (Fovea Pro 2000). Adobe PhotoshopTM is designed to edit electronic images, and is used as a platform for the Fovea Pro TM image analysis package. Fovea Pro TM functions are accessed through an extended Adobe PhotoshopTM filters menu and are capable of manipulating 8-, 16-, 24-, and 48-bit images. Although the terminology and concepts described here apply to image analysis in general, the advantage of Adobe PhotoshopTM and Fovea Pro TM is that they are cross-platform, compatible with both PC's and Macintosh computers, are widely available and relatively inexpensive. 1) Conversion of Grey Scale Images Grayscale EDS-TIFF image files are converted to RGB images to access the full range of Adobe PhotoshopTM and Fovea Pro TM functions (see image below). In Fovea Pro TM , processing operations are based on the intensity plane in the images, leaving hue and saturation levels unchanged. RGB conversion provides a false color rendition with no loss of digital information; thus, the RGB image can be directly converted back to the original 0-255 shades of gray.  2) Setting Threshold Levels Threshold values are used to make a binary image that separates 'features' from 'background'. SEM/ EDS images, however, are often poorly suited for binary transformation. In addition to elemental abundance and atomic weight (Tovey et al.,1989), image brightness can correspond to variation in particle thickness or size. Consequently, an interpretive approach to choosing a threshold value can be used, and several binary images with different thresholds may be created to distinguish among minerals of variable compositions (Tovey and Krinsley 1991; Russ 2000). Bright or white minerals in the image on the right below correspond to grains that contain predominantly Si. Grey shaded grains correspond to minerals that contian lesser amounts of Si and other elements.  3) Saturation of the Features into Binary Images Once threshold levels are set, the image must be color saturated to produce a binary image as a prerequisite for image analyses. The saturated images are used to produce mineral images and define grain shape parameters. As depicted below the binary images can be toggled between white grains on a black background or black grains on a white background depending on the type of image needed during a given image analyses procedure.  4) Setting Cutoff Values The principal limitations to grain resolution and element mapping relate to microbeam width and minimum distances of particle separation. Although the incident beam width is small, beam interaction increases the net area sampled, and decreases the spatial resolution of X-ray analysis to grains smaller than 0.5 microns (Goldstein et al.,1992). This is depicted in the image of a Monte Carlo simulation (see below right) for Si. The area sampled by the microbeam is about 5 by 8 microns. Particles smaller than 0.5 microns should not be included for analyses, and must be removed from the saturated image Depending on particle grain sizes and their distribution, the cutoff command may have to be modified; a cutoff size of less than 5 pixels eliminates most isolated areas of noise. The image on the left below is the result of setting a cutoff value and removing the smaller grains. The image above on the right corresponds to the same image prior to setting the cutoff values. 
Within grains, image brightness variations can reflect inclusions of different composition, weathering, thinner parts of the grain, or physical holes. Depending on the choice of threshold value, the variations lead to the appearance of holes in the saturated grain image. Physical Holes Physical holes can be identified by SEM secondary mode examination, and compositional differences by spot X-ray examination. Where compositional differences are consistent with known mineral compositions, the holes can be filled using the Fovea Pro TM image analysis package. Decomposition of the olivine (pictured below) has resulted in the formation of several holes. Filling the holes like this would result in a loss of information and incorrect mineral area percent calculations  Edge Effects Commonly, grain edges appear either brighter or darker than the grain, reflecting either the grain edge relief due to hardness contrast between the mineral and the epoxy, the extension of the grain beneath the epoxy, or molecular weight contrasts between adjacent grains. An example of two minerals displaying hardness contrast between the epoxy and the mineral with the resulting edge effect area highlighted with a dashed line in the following image. Mineral A= Chlorite with Si occupying the indentation, mineral B = Quartz.  Harder grains have a positive relief, whereas softer grains occupy depressions; the quality of surface polish is thus important. Sloped surfaces of mineral grains extending beneath the epoxy are detectable as beam scattering is larger than the focused incident beam (Pye and Krinsley 1984, Tovey and Hounslow 1995). The escape of backscatter electrons from the epoxy further enhances grain edge brightness (Bisdom and Schoonderbeek 1983; Dilks and Graham 1985). A Monte Carlo flight simulation of secondary and backscatter electrons for Si and Fe is depicted in the following images from Joy (1995). The Si penetration depth (image on the left) is approximately 5 microns whereas the Fe penetration depth (image on the right)is approximately 1.5 microns. The vertical line at the surface of the illustration is an approximation of the relative beam width compared to the measured volume.  Edge effects may also occur where minerals containing elements of different atomic number are juxtaposed, resulting in a transition zone where the elements of both minerals are present. In the SEM image below, mineral A) displays dark areas of plagioclase and light areas are chloritized biotite. In mineral B) the dark areas are quartz, and the white areas are epidote.  Edge effects can be addressed through application of erosion and dilation filters. Erosion removes a single pixel layer from the edge and dilation adds a single layer (Ehrlich et al.,1984). In image analysis, erosion followed by dilation is referred to as Opening, whereas dilation followed by erosion is referred to as Closing. For both, in Fovea Pro TM the pixel width can be defined with a value ranging from 1 to 99. Erosion and dilation can lead to loss of information, especially related to compositional changes associated with grain surfaces, such as secondary oxide rims. References
Adobe Photoshop TM 1999
Bisdom, E.B.A., Schoonderbeek, D. 1983
Dilks, A. and Graham, S.C. 1985
Ehrlich, R., Kennedy, S.K., Crabtree, S.J., Cannon, R.L. 1984
Fovea Pro TM 2000
Goldstein JI, Newbury DE, Echlin P, Joy DC, Romig AD Jr, Lyman CE, Fiori C and Lifshin E 1992
Joy D. 1995
Pye, K. and Krinsley, D.H. 1984
Tovey, N.K., and Hounslow, M.W., 1995
Tovey, N.K. and Krinsley, D.H. 1991
Tovey, N.K., Smart, P., Hounslow, M.W., and Leng, X.L. 1989
Russ, J.C. 2000
| |||||||||||||||||||||||||||||
