Breaking free from the matrix: Segmentation of fossil images

Abstract

The traditional method for separating a fossil from its matrix is to physically remove the matrix, risking damage to the fossil. Non-destructive technologies, such as computed tomography (CT) scanning, digitally remove the matrix. However, these techniques have multiple disadvantages: they are time consuming, involve highly skilled preparators or technicians, or require expensive equipment. Image segmenta- tion, coupled with wavelet denoising, provides a rapid, inexpensive, and non-destruc- tive way to digitally separate matrix from fossil in standard digital camera images. This method uses off-the-shelf software and produces results that can then be input into software for morphometric analysis or used to speed up more traditional segmentation techniques. A comparison of alternative methods for image segmentation and wavelet denoising indicate that Otsu’s method and other global thresholding techniques per- form best for this purpose. This approach is illustrated here with images of Eucalypto- crinites holdfasts. The resulting images are used for morphometrics. These methods have great potential for studies that utilize the rapidly growing resource of online museum digital image catalogs.