Novelty detection ※ Image Segmentation: 1. Generate random detectors McCoy devarajan 2. Apply the defectos o the sample a 3. Delete (1997) any detector misclassifying the sample data 4. Apply remaining detectors to the test image. Note pixels where a new detector responds better than any previous detector ◆ Detecting road 5. If enough pixels found improved detectors, go to Step 1 contours in aerial 6. output classified image Images ◆ Used a negative selection algorithmNovelty Detection Image Segmentation : McCoy & Devarajan (1997) Detecting road contours in aerial images Used a negative selection algorithm 1. Generate random detectors 2. Apply these detectors to the sample data 3. Delete any detector misclassifying the sample data 4. Apply remaining detectors to the test image. Note pixels where a new detector responds better than any previous detector 5. If enough pixels found improved detectors, go to Step 1 6. Output classified image