This work discusses the problem of automatic target detection in images. A study of systems and principles for automatic target recognition has been done and a number of different detection algorithms have been implemented. Both small targets and spatial extended targets are treated. Small targets are detected with an adaptive matched filter, called the RX-algorithm. Extended targets are detected with different techniques such as edge density, texture segmentation and statistical based detectors. Results from simulations on both visual and infrared images are presented. The work is a part of the development of tools for evaluation of low observables.