Attitude determination is a very important aspect in the mission of spacecraft. There are various methods of determining the attitude of a spacecraft, including, but not limited to, magnetometers, beacons and gyroscopes.[1] Star Trackers are systems consisting of one or multiple image sensors connected to a processing unit. The role of the processing unit is to detect stellar bodies and identify the patterns made by the stellar bodies in the images captured and determine the attitude of the spacecraft by comparing the patterns to those found in a pre-compiled database. This thesis describes an algorithm developed to identify stellar objects, improving a previously developed and implemented algorithm developed at the KTH Royal Institute of Technology.[2] Using a running average, the new algorithm is able to adapt to a changing background level when identifying stars. This is important as a set constant background level can become obsolete over time as the brightness of the image background changes with the movement of the spacecraft. With the use of a weighted average system, the new algorithm is able to calculate the Cartesian coordinates centers of stars within an image with sub-pixel accuracy. This improves on the 1 pixel accuracy from the old algorithm. This increase in accuracy allows for the calculation of centroids even if they occupy a small area on an image. Furthermore, an increase in the precision of the coordinates in stars may lead to an increase in the accuracy when determining attitude.