Cropping
The original image slices from the Visible Human Project male dataset are 2048 × 1216 pixels in size. However, the body does not occupy the entire area of each image. For example, there is a KODAK color bar at the bottom of each image. When we started working with the dataset, we decided to crop the images to remove the parts of the images that did not contain the body. This made it easier to perform the masking, outlined in the next section, and would reduce the amount of computational time needed to process each image and the file size of the segmented images. Initially, we cropped the images so down to 1800 × 936, where the new top left corner corresponded to (x,y) = (140,100) from the original image. This cropping worked well for the head region and was used in the partial-body heterogeneous models for AustinMan v1.0 and v1.1. However, as we continued down the body creating the masks, it became apparent that the initial cropping parameters (size and position) were removing parts of the body in the slices. Thus, we changed the cropped image size to 2048 × 1120, where the new top left corner was the same as the original top left corner. These cropping parameters were used to create the whole-body homogeneous models for Homogeneous AustinMan v1.0 and Homogeneous AustinMan v1.1.
| Previous: Methodology | Up: Methodology | Next: Masking |
