Refactoring

- Removed passing around of trivial parameters, calculating them as-needed instead.
bllanos · Jun 20, 2017 · 98ebeaf · 98ebeaf
1 parent 0f794c3
commit 98ebeaf
Show file tree

Hide file tree

Showing 12 changed files with 176 additions and 141 deletions.
diff --git a/CreateProbeDetectionModel.m b/CreateProbeDetectionModel.m
@@ -123,24 +123,23 @@
 %   created from 'probe_color_distributions_gaussian' using
 %   `mlDiscreteClassifier`. A uniform hue distribution is assumed for the
 %   image background colours.
-% - 'probe_color_distribution_increment': A scalar equal to the spacing
-%   between the samples of hue values in the range [0, 1] at which the
-%   hue density estimators have been evaluated to produce
-%   'probe_color_distributions_*'. To find the approximate value of the
-%   i-th estimator at a query value 'x' in the range [0, 1], use:
+%
+% Additionally, the output file contains the values of all parameters in
+% the first section of the script below, for reference. (Specifically,
+% those listed in `parameters_list`, which should be updated if the set of
+% parameters is changed.)
+%
+% ## Notes
+% - To find the approximate value of the i-th hue estimator at a query
+%   value 'x', where 'x' must be in the range of hues ([0, 1]), use:
 %
 %   ```
 %     queryDiscretized1DFunction(...
 %       x, probe_color_distributions_*(:, i),...
-%       probe_color_distribution_increment...
+%       hueSamplingParams(probe_color_distributions_*(:, i))...
 %     )
 %   ```
 %
-% Additionally, the output file contains the values of all parameters in
-% the first section of the script below, for reference. (Specifically,
-% those listed in `parameters_list`, which should be updated if the set of
-% parameters is changed.)
-%
 % ## References
 % - T. Gevers and H. Stokman. "Robust Histogram Construction from Color
 %   Invariants for Object Recognition". IEEE Transactions on Pattern
@@ -418,17 +417,11 @@
 G = I_double(:, :, 2);
 B = I_double(:, :, 3);
 for i = 1:n_colors
-    [...
-        probe_color_distributions_kernel(:, i),...
-        ~...
-    ] = hueVariableKernelDensityEstimator(...
+    probe_color_distributions_kernel(:, i) = hueVariableKernelDensityEstimator(...
         H, R, G, B, probe_color_masks(:, :, i),...
         rgb_sigma_polyfit, probe_color_distribution_resolution...
     );
-    [...
-        probe_color_distributions_gaussian(:, i),...
-        probe_color_distribution_increment...
-    ] = hueGaussianDensityEstimator(...
+    probe_color_distributions_gaussian(:, i) = hueGaussianDensityEstimator(...
         H, probe_color_masks(:, :, i), probe_color_distribution_resolution...
     );
 end
@@ -439,35 +432,34 @@
         legend_names{i} = sprintf('Probe color %d', i);
     end
     plotHueDensityEstimator(...
-        probe_color_distribution_increment, probe_color_distributions_kernel, legend_names...
+        probe_color_distributions_kernel, legend_names...
     );
     title('Hue variable kernel density estimators for colors on the probe')
     plotHueDensityEstimator(...
-        probe_color_distribution_increment, probe_color_distributions_gaussian, legend_names...
+        probe_color_distributions_gaussian, legend_names...
     );
     title('Hue Gaussian density estimators for colors on the probe')
 end
 
 %% Create a hue classifier
 
+h_inc = hueSamplingParams( probe_color_distribution_resolution );
 probe_color_classifier_kernel = mlDiscreteClassifier(...
-    probe_color_distributions_kernel, probe_color_distribution_increment,...
-    'periodic'...
+    probe_color_distributions_kernel, h_inc, 'periodic'...
 );
 
 probe_color_classifier_gaussian = mlDiscreteClassifier(...
-    probe_color_distributions_gaussian, probe_color_distribution_increment,...
-    'periodic'...
+    probe_color_distributions_gaussian, h_inc, 'periodic'...
 );
 
 if plot_hue_classifiers
     plotHueClassifier(...
-        probe_color_distribution_increment, probe_color_classifier_kernel,...
+        probe_color_classifier_kernel,...
         n_colors...
     );
     title('Hue variable kernel distribution classifier for colors on the probe')
     plotHueClassifier(...
-        probe_color_distribution_increment, probe_color_classifier_gaussian,...
+        probe_color_classifier_gaussian,...
         n_colors...
     );
     title('Hue Gaussian distribution classifier for colors on the probe')
@@ -479,8 +471,7 @@
         'probe_color_distributions_kernel',...
         'probe_color_classifier_kernel',...
         'probe_color_distributions_gaussian',...
-        'probe_color_classifier_gaussian',...
-        'probe_color_distribution_increment'...
+        'probe_color_classifier_gaussian'...
     } ];
 uisave(save_variables_list,'probeDetectionModel')
 disp('Reminder: The output model is specific to the probe, camera, and camera parameters.')
diff --git a/DetectProbe.m b/DetectProbe.m
@@ -216,10 +216,8 @@
 
 %% Load the probe detection model
 model_variables_required = {...
-        'probe_color_distribution_resolution',...
         'probe',...
         'probe_color_distributions_kernel',...
-        'probe_color_distribution_increment',...
         'model_filename'...
     };
 load(detection_model_filename, model_variables_required{:});
@@ -244,7 +242,6 @@
 else
     probe_mask_initial = detectBoundingBoxes(...
         I, probe,...
-        probe_color_distribution_resolution,...
         probe_color_distributions_kernel, rgb_sigma_polyfit,...
         detectBoundingBoxesParams, detectBoundingBoxesVerbose...
     );
@@ -255,7 +252,7 @@
 [...
     interest_points_detected, probe_regions_bw_final_filtered...
 ] = detectWithinBoundingBox(...
-    I, probe, probe_mask_initial, probe_color_distribution_resolution,...
+    I, probe, probe_mask_initial,...
     probe_color_distributions_kernel, rgb_sigma_polyfit,...
     detectWithinBoundingBoxParams, detectWithinBoundingBoxVerbose...
     );
@@ -300,7 +297,7 @@
 [...
     probe_colors_detected_left, probe_colors_detected_right...
 ] = labelDetectedColors(...
-    I, probe_color_distributions_kernel, probe_color_distribution_increment,...
+    I, probe_color_distributions_kernel,...
     probe_regions_bw_final_filtered,...
     model_from_image_detected_pca_space, model_from_image_detected,...
     model_from_image_transform, model_from_image_detected_axes(1, :),...

diff --git a/detectBoundingBoxes.m b/detectBoundingBoxes.m
@@ -1,6 +1,5 @@
 function [ mask ] = detectBoundingBoxes(...
     I, probe,...
-    probe_color_distribution_resolution,...
     probe_color_distributions, rgb_sigma_polyfit,...
     params, varargin...
     )
@@ -9,15 +8,13 @@
 % ## Syntax
 % mask = detectBoundingBoxes(...
 %     I, probe,...
-%     probe_color_distribution_resolution,...
 %     probe_color_distributions, rgb_sigma_polyfit,...
 %     params [, verbose]...
 % )
 %
 % ## Description
 % mask = detectBoundingBoxes(...
 %     I, probe,...
-%     probe_color_distribution_resolution,...
 %     probe_color_distributions, rgb_sigma_polyfit,...
 %     params [, verbose]...
 % )
@@ -38,11 +35,6 @@
 %   Refer to the documentation of './CreateProbeDetectionModel.m' for
 %   details.
 %
-% probe_color_distribution_resolution -- Probe colour estimator sample count
-%   The number of equally-spaced samples in the range of hue values from 0
-%   (inclusive) to 1 (inclusive) at which the density estimators for probe
-%   colour bands have been evaluated.
-%
 % probe_color_distributions -- Probe colour estimators
 %   Discretized density estimators of image hue values corresponding to the
 %   different coloured bands on the probe, in the same order (starting from
@@ -106,7 +98,7 @@
 %% Parse input arguments
 
 nargoutchk(1,1);
-narginchk(6,7);
+narginchk(5,6);
 
 if ~isempty(varargin)
     verbose = varargin{1};
@@ -144,7 +136,7 @@
 %% Find an initial bounding area for the probe
 
 [ probe_regions_initial, probe_regions_bw_initial] = extractBinaryRegions(...
-    I, probe_color_distributions, probe_color_distribution_resolution,...
+    I, probe_color_distributions,...
     rgb_sigma_polyfit,...
     params.erosion_radius,...
     [],...

diff --git a/detectWithinBoundingBox.m b/detectWithinBoundingBox.m
@@ -1,29 +1,29 @@
 function [...
     interest_points, probe_regions_bw_filtered...
     ] = detectWithinBoundingBox(...
-        I, probe, mask, probe_color_distribution_resolution,...
+        I, probe, mask,...
         probe_color_distributions, rgb_sigma_polyfit,...
         params, varargin...
     )
 %DETECTWITHINBOUNDINGBOX Find the probe within a bounding region
 %
 % ## Syntax
 % [ interest_points ] = detectWithinBoundingBox(...
-%       I, probe, mask, probe_color_distribution_resolution,...
+%       I, probe, mask,...
 %       probe_color_distributions, rgb_sigma_polyfit,...
 %       params [, verbose]...
 % )
 % [...
 %     interest_points, probe_regions_bw_filtered...
 % ] = detectWithinBoundingBox(...
-%       I, probe, mask, probe_color_distribution_resolution,...
+%       I, probe, mask,...
 %       probe_color_distributions, rgb_sigma_polyfit,...
 %       params [, verbose]...
 % )
 %
 % ## Description
 % [ interest_points ] = detectWithinBoundingBox(...
-%       I, probe, mask, probe_color_distribution_resolution,...
+%       I, probe, mask,...
 %       probe_color_distributions, rgb_sigma_polyfit,...
 %       params [, verbose]...
 % )
@@ -33,7 +33,7 @@
 % [...
 %     interest_points, probe_regions_bw_filtered...
 % ] = detectWithinBoundingBox(...
-%       I, probe, mask, probe_color_distribution_resolution,...
+%       I, probe, mask,...
 %       probe_color_distributions, rgb_sigma_polyfit,...
 %       params [, verbose]...
 % )
@@ -54,11 +54,6 @@
 %   channel, not three), where the nonzero pixels ideally define regions
 %   containing the colour bands of the probe.
 %
-% probe_color_distribution_resolution -- Probe colour estimator sample count
-%   The number of equally-spaced samples in the range of hue values from 0
-%   (inclusive) to 1 (inclusive) at which the density estimators for probe
-%   colour bands have been evaluated.
-%
 % probe_color_distributions -- Probe colour estimators
 %   Discretized density estimators of image hue values corresponding to the
 %   different coloured bands on the probe, in the same order (starting from
@@ -138,7 +133,7 @@
 %% Parse input arguments
 
 nargoutchk(1,2);
-narginchk(7,8);
+narginchk(6,7);
 
 if ~isempty(varargin)
     verbose = varargin{1};
@@ -173,7 +168,7 @@
 %% Find final regions corresponding to probe colours
 
 [ probe_regions_final, probe_regions_bw_final] = extractBinaryRegions(...
-    I, probe_color_distributions, probe_color_distribution_resolution,...
+    I, probe_color_distributions,...
     rgb_sigma_polyfit,...
     params.erosion_radius,...
     mask,...
@@ -204,9 +199,10 @@
 
 if display_regions_colored
     probe_regions_bw_final_display = zeros(image_height, image_width, image_n_channels);
+    [~, color_distribution_resolution] = hueSamplingParams( probe_color_distributions(:, 1) );
     for i = 1:n_colors
         [ ~, peak_hue_index ] = max(probe_color_distributions(:, i));
-        peak_hue = (peak_hue_index - 1) / (probe_color_distribution_resolution - 1);
+        peak_hue = (peak_hue_index - 1) / (color_distribution_resolution - 1);
         peak_rgb = hsv2rgb([peak_hue, 1, 1]);
         probe_regions_bw_final_filtered_i = probe_regions_bw_filtered(:, :, i);
         probe_regions_bw_final_display = probe_regions_bw_final_display +...