Abstarct: Recent technological advances have made manufacturing lightweight and head-mountable cameras with many applications, possible. Google Glass, as a wearable gadget, is a publicly affordable example. These gadgets have introduced the idea of egocentric (first person) vision as a new frxamework for the computer vision community. In this frxamework, one is capable to recognize objects and activities in his own field of view. Vision capabilities are also augmented in medical surgeries, industry and art, thus new solutions to overcome challenges in these fields are introduced. Nonlinear motion, blur, fast transition and runtime of processing algorithms are such challenges. In this thesis, to optimize a superpixel algorithm in the frxamework of egocentric vision, a new approach is proposed. In the frxamework of egocentric videos captured from the wearable gadgets, a fast and portable implementation of a processing algorithm is needed. In order to get closer to the real-time performance, we have established a trade-off between segmentation quality and runtime of a state-of-the-art algorithm in the proposed method. For a video of 1280×720 resolution, we have achieved a 23% up to 64% increase in performance for various approaches. The algorithm optimization, has caused a slight decrease in segmentation quality in homogenous regions of a video frxame. This decrease however, has not influenced the edge segmentation quality.