A multi-class bi-level thresholding method for accurate anthropometric measurements of scanned plantar images

Traditionally, the anthropometric measurements of foot or insole were taken manually by an anthropometrist for clinical and sporty purposes. Digital foot scanners have been developed in recent years to yield anthropometrists digital image of plantar with pressure distribution and anthropometric information. In this paper, an automated thresholding method based on Shanbag entropy and the weighted gray level spatial correlation histogram is presented for analysis of scanned foot images. The presented method compared with traditional two-dimensional histogram and other single-class-based methods automatically accurately detects the foot edges and areas. Also, by background removal, it attains a perfect foot image while in which white pixels of the resulting thresholded image correspond to the points of the scanner which detects pressure and on the contrary the black pixels correspond to background. We are taking into accounts the image’s local properties along with its global properties in a fuzzy domain by employing weighted gray level spatial correlation histogram (W-GLSC). First, a three-dimensional histogram based on the statistics of the gray levels’ probability and similarity with neighboring pixels (GLSC histogram) is weighted by the membership values assigned by Jawahar clustering method for foreground–background discrimination. Then, the Shanbag entropy is used to maximize the information transfer from the original image to the resulting thresholded image. Resulting binary images are undergone anthropometric measurements by considering the scale factor of pixel size to metric scale. The proposed method is finally applied to plantar images obtained through scanning feet of randomly selected subjects by a foot scanner system as our experimental setup described in the paper.