Human Face Image Segmentation using Level Set Methodology

Abstract

Face segmentation plays an important role in various applications such as human computer interaction, video surveillance, biometric systems, and face recognition for purposes including authentication and authorization. The accuracy of face classification system depends on the correctness of segmentation. Robustness of the face classification system is determined by the segmentation algorithm used, and the effectiveness in segmenting images of similar kind. This paper explains the level set based segmentation for human face images. The process is done in two stages: In order to get better accuracy, binarization of the image to be segmented is performed. Next, segmentation is applied on the image. Binarization is the process of setting pixel intensity values greater than some threshold value to "on" and the rest to "off". This process converts the input image into binary image which is used for segmentation. Second process is image segmentation for eliminating the background portion from the binarized image which is obtained after the binarization of the original image. Conventional approaches use separate methods for binarization and segmentation. In this paper we investigate the use of recently introduced convex optimization methods, selective local/global segmentation (SLGS) algorithm [16] for simultaneous binarization and segmentation. The approach is tested in MATLAB and satisfactory results were obtained.

Human Face Recognition under Varying Illumination Condition Using Wavelet Transform

“…To overcome the problem of noise in the image, de-noising method [4] must be processed. De-noising is a method which is used to remove the unwanted noise from an image.…”

Dot Area Coverage Percentage Measurement Method for Plate Microscopic Image Based on Color Segmentation

“…The hue of plates using in industrial production can be completely described by the three primary colors (red, green, blue) and its three mixed color (yellow, cyan, and magenta) in HSI color space. Therefore, according to hue, we can divide the hue of HSI color space into six parts, red, yellow, green, cyan, blue, and magenta [10], and the corresponding hue Angle is(330…”

Review of topology optimisation refinement processes for sheet metal manufacturing in the automotive industry

“…For some optimisation solvers it is therefore not possible to avoid variable densities and so the "issue" must be addressed in the post-processing stage. More recent optimisation processes include Level-Set topology optimisation (Challis 2009;Kumaravel, M. et al 2012;Liu 2015) which creates a binary model of either full densities or no density (1s and 0s only) from variable density optimisation results. This effectively cleans up the "grey" edges of a model, creating clean lines better defining structural boundaries.…”