For successful SLAM, perception of the environment is important. This paper proposes a scheme to autonomously detect features which are used as natural landmarks for indoor SLAM. Features are roughly selected by using entropy maps which measure the level of randomness of information. The selected features are evaluated by the saliency map based on similarity maps which measure the level of similarity between the selected features and the given image. In the saliency map, it is possible to distinguish the salient features from the background. In this research, the HSV color space is adopted for color representation instead of the RGB space. The robot estimates its pose using the detected features and builds a grid map of the unknown environment using a range sensor. The feature positions are stored in the grid map. Experimental results show that the feature detection proposed in this paper can autonomously detect features in unknown environments reasonably well.