Typos.

Add README.
Rewrite global RX.

README.md

@@ -0,0 +1,55 @@
+Color Anomaly Detection
+=======================
+
+Written as part of a class project on color anomaly detection for search and rescue purposes, this repository contains a number of Matlab implementations of common anomaly detection algorithm from literature on hyperspectral techniques, as well as our own algorithm (PCAG) selected to balance performance and computation time.
+
+The full technical report on the project, including citations and evaluations of the different anomaly detection algorithms is available online at:
+
+[Color Outlier Detection for Search and Rescue](http://www.nathanntg.com/writing/color-anomaly-detection.pdf)
+
+
+Implemented hyperspectral algorithms
+------------------------------------
+
+The following hyperspectral algorithms are implemented in this repository:
+
+* Global RX (`rxg`)
+* Local RX (`rxl`)
+* Dual Window-based Eigen Separation Transform (`dwest`)
+* Nested Spatial Window-based Target Detection (`nswtd`)
+* Multiple Window Nested Spatial Window-based Target Detection (`mwnswtd`)
+
+Each function above takes a single argument representing the image as a Matlab double matrix with spectral intensities between 0 and 1.
+
+Novel anomaly detection algorithms
+----------------------------------
+
+The following new algorithms are implemented in this repository:
+
+* Principal Component Analysis Gaps (`pcag`)
+* Multiple Window Principal Component Analysis Gaps (`mwpcag`)
+
+Utility functions
+-----------------
+
+A few useful utility functions are implemented in this repository:
+
+* `im_norm = normalize_image(im)` normalizes an image to be stored in a Matlab double matrix and discarding any transparency channel data.
+* `[scene, target] = build_scene(file, num_anomalies, blended)` superimposes random anomalies (from an "anomalies" directory) onto a scene (from a "scenes" directory) in a random position and rotation, and with luminance matching to the surrounding pixels. Returns both the new scene (scene) and a target image (target), representing anomaly positions.
+* `[tpr, fpr, th, auc] = roc_anomaly(target, out)` calculates the ROC curve for a particular anomaly detector output (out) based on the target image (target). Shows a plot and returns the true-positive rate (tpr), false-positive rate (fpr), thresholds (th) and area under the curve (auc).
+
+Analysis scripts
+----------------
+
+* `run` evaluates the above algorithms over a number of scenes and color spaces in parallel (using `run_script`)
+* `analyze` evaluates the output of the run function above across algorithms and colorspaces
+* `analyze_environ` evaluates the output of the run function above across scene types (assuming consistently prefixed scene names)
+
+### Authors
+
+**L. Nathan Perkins**
+
+- <https://github.com/nathanntg>
+- <http://www.nathanntg.com>
+
+**Travis Marshall**

analyze.m

@@ -48,7 +48,7 @@
 		end
 
 		% calculate AUC
-		[~, ~, ~, auc] = roc_anomally(target, out);
+		[~, ~, ~, auc] = roc_anomaly(target, out);
         title(sprintf('%s with %s color space', algos_nice{j}, color_spaces{i}));
 		print(gcf, sprintf('roc/%d-%s.png', i, algo), '-dpng', '-r300');
 

analyze_environ.m

@@ -53,7 +53,7 @@
 		end
 
 		% calculate AUC
-		[~, ~, ~, auc] = roc_anomally(target, out);
+		[~, ~, ~, auc] = roc_anomaly(target, out);
         title(sprintf('%s for %s scenes', algos_nice{j}, scene_nice{i}));
 		print(gcf, sprintf('roc/%s-%s.png', scene_nice{i}, algo), '-dpng', '-r300');
 

pcag_block.m

@@ -1,7 +1,7 @@
 function px = pcag_block(block_struct)
 %PCAG Perform the PCAG algorithm on img.
 
-% size of anomally
+% size of anomaly
 min_size = 4;
 max_size = min(100, floor(0.5 * prod(block_struct.blockSize)));
 

performance.m

@@ -1,5 +1,5 @@
 % algorithms to performance test
-algos = {@RX_global, @rxl, @dwest, @nswtd, @mwnswtd, @pcag, @mwpcag, @pcad, @knna};
+algos = {@rxg, @rxl, @dwest, @nswtd, @mwnswtd, @pcag, @mwpcag, @pcad, @knna};
 algos_nice = {'Global RX', 'Local RX', 'DWEST', 'NSWTD', 'MW-NSWTD', 'PCAG', 'MW-PCAG', 'PCAD', 'KNN'};
 
 % scenes to compare

roc_anomally.m → roc_anomaly.m

@@ -1,6 +1,6 @@
-function [tpr, fpr, th, auc] = roc_anomally(im_target, im_out)
-%ROC_ANOMALLY Plot receiver operating characteristic and return values.
-%   Takes a target image and the output of an anomally detector.
+function [tpr, fpr, th, auc] = roc_anomaly(im_target, im_out)
+%ROC_ANOMALY Plot receiver operating characteristic and return values.
+%   Takes a target image and the output of an anomaly detector.
 %   Shows a plot of the ROC curve.
 
 % remove color component

run_script.m

@@ -51,7 +51,7 @@ function run_script(scene_file)
     % RX GLOBAL
     fname = sprintf('output/%s-%d-rx.mat', scene_file, j);
     if ~exist(fname, 'file')
-        img_rx = RX_global(img);
+        img_rx = rxg(img);
         save(fname, 'img_rx');
     end
 

rxg.m

@@ -0,0 +1,29 @@
+function im_result = rxg(img)
+% Perform RX algorithm on global of image
+
+% Dimensions of block
+[height, width, channels] = size(img);
+
+% Separate block into color channels to calculate K matrix
+cc = reshape(img, [], channels);
+k_inv = inv(cov(cc));
+
+% Create mean color column vector
+mu = mean(cc);
+
+% zero-mean
+cc_rows = size(cc, 1);
+cc = cc - (ones(cc_rows, 1) * mu);
+
+% Preallocate
+im_result = zeros(cc_rows, 1);
+
+for i = 1:cc_rows
+    % Locate center pixel and convert to column vector
+    r = cc(i, :);
+
+    % Run RX detector on center pixel
+    im_result(i) = r * k_inv * r';
+end
+
+im_result = reshape(im_result, height, width);