Recognition, segmentation and retrieval of gang graffiti images on a mobile device

In this paper we describe three methods for recognition, segmentation and retrieval of gang graffiti images. The first method is color recognition based on touchscreen tracing, the second method is color image segmentation based on Gaussian thresholding and the third method is content based image retrieval. Our experimental results show an image retrieval accuracy of 92.8% for gang graffiti scene recognition and an image retrieval accuracy of 50.0% for gang graffiti component classification. The experiments also show an average image retrieval time of 0.56 seconds, from which the scoring process takes on average 984 microseconds.