Anomalous Pixels
Features:
Identifies anomalous pixels in the 3D datasets using Reed-Xialoi (RX) detection algorithm.
Steps:
1. Select Image Statistics → Anomalous pixels.
2. Enter Confidence coefficient to set a threshold at which any pixels with the higher confidence coefficient will be considered anomalous. The default level is 0.99. This threshold represents the RX score above which a pixel is an anomaly with 99.0 % confidence.
3. As the first step, we suggest using the default value and press Apply. You can identify a more suitable Confidence coefficient value by analyzing the Plot of the cumulative probability distribution of RX score values.
IDCubePro® will compute the plot of the Cumulative probability distribution of RX score values (scale from 0 to 1). Pixels with a high RX score are likely anomalous pixels. IDCubePro® also displays the RX score map where each pixel in the dataset is assigned the raw RX score. Both graphs appear in a pop-up panel Reed-Xialoi (RX) Scores side-by-side.
Tip 1: Use the Strip Toolbar from the right upper corner of the graph (to see this toolbar, hover over the right upper corner of the graph) to identify the threshold. Enter this new threshold value into Confidence coefficient field to tune the anomalous pixel detection.
Tip 2: Right-click on the Colorbar to change the color scheme or adjust the colors using an interactive colorbar shift.
IDCubePro® automatically applies a threshold criterion to detect anomalous pixels with RX score greater than the computed threshold. The result is a binary image in which the anomalous pixels are assigned the intensity value 1 and other pixels are assigned 0. This binary image is generated in a second pop-up panel named Detected Anomalous Pixels (binary). The detected anomalous pixels are overlaid with the automatically generated pseudo RGB image shown in Overlaid Image panel.
Notice that many anomalous pixels are found in the coin. These anomalous pixels correspond to the highly reflective areas of the coin’s surface.
Additional Information:
The method detects anomalous pixels in the hyperspectral data using the Reed-Xialoi (RX) detector. The RX detector calculates a score for each pixel as the Mahalanobis distance between the pixel and the background. The background is characterized by the spectral mean and covariance of the datacube.
References:
[1] Reed, I.S., and X. Yu. “Adaptive Multiple-Band CFAR Detection of an Optical Pattern with Unknown Spectral Distribution.” IEEE Transactions on Acoustics, Speech, and Signal Processing 38, no. 10 (October 1990): 1760–70. https://doi.org/10.1109/29.60107.
[2] Chein-I Chang and Shao-Shan Chiang. “Anomaly Detection and Classification for Hyperspectral Imagery.” IEEE Transactions on Geoscience and Remote Sensing 40, no. 6 (June 2002): 1314–25. https://doi.org/10.1109/TGRS.2002.800280.
