Thesis (M.S., Computer Science)--University of Idaho, June 2014 | Image Processing and computer vision has always been a major component in the field of robotics and machine learning. The information extracted from image processing varies widely depending upon the nature of tasks to be performed by the robot. The goal of this research is to enable a robot to find a path for autonomous trail following. In the research an evolutionary approach has been used to determine the minimal regions in input images, which are discriminating enough for the robot to distinguish between path and non-path. Image processing is computationally intensive in nature, requiring significant processing time and main memory. So, any reduction in area of the image to be processed for decision making is always advantageous. In this research two sets of experiments are done to compare the effectiveness of evolved regions for image processing against processing the whole image. In the first set the robot processed the entire frame from the camera to make decisions for driving on a trail by avoiding non-trail. In the second set image processing is done on only the evolved region of the frame to make the decision. It was found that image processing over the evolved region covering less than 30% of the captured image can give results comparable to processing the whole image.