Automatic Image Orientation Detection Using the Supervised Self-organizing Map

Automatic detection and correction of image orientation is of great importance in intelligent image processing. In this paper, we present an automatic image orientation detection algorithm based on the supervised selforganizing map (SOM). The SOM is trained by using compact and efficient low-level chrominance (color) features in a supervised manner. Experiments have been conducted on a database containing various images with different compositions, locations, and contents. The proposed algorithm achieves an accuracy of 75% using the SOM trained by 600 images. In comparison with three peer systems, the proposed system achieves decent accuracy with the compact feature vector, the minimum training time, and the minimum training data. This framework will bridge the gap between computer and human vision systems and is applicable to other problems involving semantic content understanding.