IODP Proceedings    Volume contents     Search
iodp logo



Cutting and scaling the image

The initial step in the procedure to build a compound image of an entire core is to extract from the digital picture of a core section only that part of the image that is sediment. (We use the term “compound” to avoid “composite,” which is associated with the composite depth scale.) Digital core scan images of a 1.5 m long section are ~16,000 pixels × 860 pixels, whereas the core image itself is on the order of 15,000 pixels × 700 pixels within the image. At 15,000 pixels per 1.5 m long section, the resolution of the digital image is about 0.1 mm. One end of a typical section image is displayed in Figure F1. Note the meter rule across the top of image supplying scale. We use the white cross-hair cursor shown in the figure to mark the top of the cored material. In general, as long as the top of the cored material is within 1 cm of the top of the core tube we used 0 cm as the top level of material. A similar procedure is carried out for the lower end of the section, with a provision to mark the extent of missing material because of water content samples taken from the bottom of sections prior to splitting. The section length identified through the cursor and meter rule is checked against the coring table downloaded from the IODP database. Typically, as with the top of the section, the section length in the database is within 1 cm of the length seen in the image.

Once the horizontal positions of the top and bottom of the sediment in the image have been delineated, a 400 pixel wide swath is cut from the image centered around pixel 500 (see the vertical axis in Fig. F1). The data within this image are then scaled in meters below seafloor (mbsf), the depth to the top of the section extracted from the coring summary. During the scaling process, the subimage is interpolated to a common scaling factor of 0.35 mm per pixel along the depth axis. This scaling factor is ~⅓ that of the original data and is a trade-off between maximum resolution and the desire to keep the composite core images to a manageable size. Once each section of core has been processed, individual sections are concatenated into a compound core image (Fig. F2). In the figure we have also brightened the image by expanding the color ranges. Image enhancement brings out more subtle layering than is visible in the original images and even the cores themselves.

The process of assembling a compound core image is fast and simple. It takes <5 min to process a seven section core and produce the scaled compound image. It is then a simple exercise to plot other sets of data over the image, such as magnetic susceptibility from the multisensor track (MST) (Fig. F3). The scanner used during Expedition 303 was a little too slow to allow these types of plots to be done before sampling and core description was complete, but newer, faster scanning equipment will certainly allow plots such as those in Figure F3 to be available as individual sections are described.

Hole and site composites

Compound core images were initially calibrated to mbsf values but were easily shifted to mcd using the affine table generated by Splicer, the core correlation program used to develop the composite depth scale. After each core was scaled to mcd, the Splicer splice table was used to copy from compound core images those intervals of the images that were part of the composite site. The subimages were then assembled to produce a composite site image using the mcd scale. Splicer-generated data files from the MST can be overlain on the composite image as well as discrete data such as carbonate content (Fig. F4).

Color images are stored in the computer as three-dimensional matrixes. Raw data from the Expedition 303 core scanner were stored as RGB layers. Our software allowed us to transform the RGB images to HSL images as an alternative color image storage protocol. Profiles of R, G, B, H, S, and L were extracted from the composite site images to facilitate comparison of image data with other discrete data. Each profile represents the average value of the 20 center pixels at each depth point. The carbonate content data displayed over the composite core image in Figure F4 suggest a relationship between higher carbonate values and lighter core colors. We plotted the same carbonate content values along with a profile of L from the Site U1304 composite core image in Figure F5. Agreement between the two sets of data is apparent and is further illustrated in a cross-plot of carbonate content and corresponding value of L in Figure F6. The majority of the anomalous points in Figure F6 are associated with thick diatomite intervals seen deeper at the site.

Core to composite correlation

We displayed each compound core image from Site U1304 versus its assigned mcd depth in Figure F7. A close-up of only the top 25 mcd is shown in Figure F8. By plotting the tops and bottoms of the intervals used to make up the mcd site, it is possible to see how well intervals of core outside the composite agree with those used to form it. For Site U1304 it can be seen in Figure F7 that the composite comes almost exclusively from Holes U1304A and U1304B. The mismatch of Core 303-U1304C-2H is obvious in Figure F8.

Several studies have examined distortion or natural variability core-to-core in deep-sea drilling sediments in the past (e.g., Hagelberg et al., 1995; Pälike et al., 2005). They found that differences in mcd depths of features based on profile data such as MST or color scanners generally amounted to <20 cm or so. We performed the same analysis on five sites from Expedition 303 using both profile data and our compound and composite images. Our technique is illustrated in Figure F9. Initially, a compound core image is plotted next to the composite site image. Cursors are used to identify and record tie points between the compound and composite images. Any of the profile data, such as magnetic susceptibility in Figure F9, can be overlain on the image data to aid in choosing tie points. Once an initial set of tie points has been chosen, an interpolated image of the compound core is generated that has been either shrunk or stretched to match the spacing of the tie points. The new image is then plotted next to the composite image. We found that the use of an interpolated image was a great help in refining the placing of tie points and that it was often the case that centimeter-scale layering could be perfectly matched between compound and composite core images. The tie points used in Figure F9 are shown in Table T1.