Merge branch '64-dichotomie-ajout-du-module-dichotomy-py' into 'release'

feat: add dichotomy module in refinement package See merge request 3d/PandoraBox/pandora2d!71
CNES · Feb 9, 2024 · 4e2739e · 4e2739e
2 parents 1875185 + eab2b2c
commit 4e2739e
Show file tree

Hide file tree

Showing 8 changed files with 598 additions and 45 deletions.
diff --git a/docs/source/userguide/step_by_step/refinement.rst b/docs/source/userguide/step_by_step/refinement.rst
@@ -7,21 +7,57 @@ Theoretical basics
 ------------------
 The purpose of this step is to refine the disparity identified in the previous step.
 
-The available refinement method is :
+The available refinement methods are:
+    * **Interpolation**:
 
-* **Interpolation** : It consists on 3 differents steps:
+        It consists on 3 different steps:
+            * First, the cost_volumes is reshaped to obtain the 2D (disp_row, disp_col) costs map for each pixel, so we will obtain (row * col) 2D cost maps.
+            * The cost map of each pixel is interpolated using scipy to obtain a continuous function.
+            * Then, the interpolated functions are minimized using scipy to obtain the refined disparities.
 
-    * First, the cost_volumes is reshaped to obtain the 2D (disp_row, disp_col) costs map for each pixel, so we will obtain (row * col) 2D cost maps.
-    * The cost map of each pixel is interpolated using scipy to obtain a continuous function.
-    * Then, the interpolated functions are minimized using scipy to obtain the refined disparities.
+    * **Dichotomy**:
+
+        It’s an iterative process that will, at each iteration:
+            * compute the half way positions between each best candidate in the cost volume and its nearest neighbours.
+            * compute the similarity coefficients at those positions using the given filter method.
+            * find the new best candidate from those computed coefficients.
 
 Configuration and parameters
 ----------------------------
-+---------------------+-------------------+--------+---------------+---------------------+----------+
-| Name                | Description       | Type   | Default value | Available value     | Required |
-+=====================+===================+========+===============+=====================+==========+
-| *refinement_method* | Refinement method | string |               |"interpolation"      | No       |
-+---------------------+-------------------+--------+---------------+---------------------+----------+
+
+.. list-table:: Configuration and parameters
+    :header-rows: 1
+
+    * - Name
+      - Description
+      - Type
+      - Default value
+      - Available value
+      - Required
+    * - *refinemement_method*
+      - Refinement method
+      - string
+      -
+      - | "interpolation",
+        | "dichotomy",
+      - Yes
+    * - *iterations*
+      - Number of iterations
+      - integer
+      -
+      - | 1 to 9
+        | *if above, will be bound to 9*
+        | **Only available if "dichotomy" method**
+      - Yes
+    * - *filter*
+      - Name of the filter to use
+      - str
+      -
+      - | "sinc",
+        | "bicubic",
+        | "spline",
+        | **Only available if "dichotomy" method**
+      - Yes
 
 **Example**
 

diff --git a/notebooks/introduction_and_basic_usage.ipynb b/notebooks/introduction_and_basic_usage.ipynb
@@ -131,11 +131,11 @@
    "source": [
     "### Available implementations for each step\n",
     "\n",
-    "| Step                      | Algorithms implemented |\n",
-    "|:--------------------------|:-----------------------|\n",
-    "| Matching cost computation | SAD / SSD / ZNNC       |\n",
-    "| Disparity computation     | Winner-Takes-All        |\n",
-    "| Subpixel disparity refinement | Interpolation      |"
+    "| Step                          | Algorithms implemented    |\n",
+    "|:------------------------------|:--------------------------|\n",
+    "| Matching cost computation     | SAD / SSD / ZNNC          |\n",
+    "| Disparity computation         | Winner-Takes-All          |\n",
+    "| Subpixel disparity refinement | Interpolation / Dichotomy |"
    ]
   },
   {