Human fingernail segmentation

In this paper we present a method for segmentation of fingernail patterns and differentiate them as distinct nail parts; fingernail plate with lunula and distal free edge of nail plate. In the research work, focus is on fixed area of the fingernail plate plus lunula, as it remains unchanged in structure, where as the distal nail edge extends and changes in structure over a period of time. In order to segment fingernail parts, we have devised an algorithm that automatically separates unchanging region of fingernail plate from free distal edge of nail structure. The fingernail plate that includes lunula within (may or may not be prominently present in fingernails), is used as biometric in our advance study. Theory suggests, every fingernail within finger formation comprises of the brightest regions amongst the captured finger data set (in our system). Proposed method is of two stages. In first stage, color image is converted to gray scale and contrast enhancement is applied using adaptive histogram equalization. In second stage, we perform segmentation using watershed method that exercises maxima and minima properties of marker controlled watershed principles. In order to verify the results of the algorithm, we have constructed a confusion matrix where evaluation has been done with ground truth. Additionally, the segmented object's from both the methods is considered for quality metrics assessment. Similarity accuracy between the ground truth and watershed result is 84.0% correctness for fingernail plate. Initial fingernail segmentation results are promising, supporting its use for biometric application.