A Review of Computer Vision Segmentation Algorithms

1 Introduction The remote sensing and computer vision communities share a common goal of extracting useful information from raw imagery. Both communities have exploited several trends that support the increasingly timely, cost effective, accurate, and effective automated extraction of information from raw imagery that include increasingly powerful, affordable, and available computer hardware; increasingly sophisticated software tools, both commercial and open source, that have a proven track record; a growing community of computer knowledgeable users; and an increasing proliferation of sophisticated sensors, both active and passive, ranging from handheld digital cameras to commercial satellites with drastically improved spatial, radiometric, and temporal resolution. While the computer vision (CV) community has many of the same goals as the remote sensing (RS) cummunity, its applications are not focused in characterising the earth's surface but include a much wider range of applications. and entertainment. Computer vision systems have been designed that can control robots or autonomous vehicles, inspect machine parts, detect and recognize human faces, retrieve images from large databases according to content, reconstruct large objects or portions of cities from multiple photographs, track suspicious people or objects in videos, and more. Remote sensing systems work on multi-spectral images that captures image data at specific frequencies across the electromagnetic spectrum. These images contain multiple bands, some from light frequencies visible to humans and some from frequencies beyond the visible light range, such as in-frared. Computer vision systems have been developed for all types of images, but mainly work with single-band graytone images, three-band color images, and single-band depth images, sometimes registered to color images of the same scene. In spite of this difference, many of the same tools can be used in both computer vision and remote sensing. Table 1 shows a simplified classification of computer vision tools, which are available in both open-source, such as Intel's OpenCV library in C++ [15] and NIH's ImageJ library in Java [3], and 2 Basic Tools Example Applications filters noise suppression edge detection texture description segmentation object recognition image retrieval medical image analysis interest operators image matching motion analysis object recognition image retrieval photogrammetric operations 3D reconstruction Table 1: A Simple Classification of Computer Vision Tools commercial packages such as Matlab with its Image Processing and Signal Processing toolkits [14]. Filter tools, which should be familiar to most RS workers, are used to enhance images or extract low-level features. Sharpening, brightening, noise removal, edge detection, and texture feature …