Traffic sign recognition is a difficult task if we aim at detecting and recognizing signs in images captured from unfavorable environments. Complex background, weather, shadow, and other lighting-related problems may make it difficult to detect and recognize signs in the rural as well as the urban areas. We employ discrete cosine transform and singular value decomposition for extracting features that defy external disturbances, and compare different designs of detection and classification systems for the task. Experimental results show that our pilot systems offer satisfactory performance when tested with very challenging data.