summaryrefslogtreecommitdiff
path: root/utils/raspberrypi/ctt/ctt_dots_locator.py
diff options
context:
space:
mode:
authorBen Benson <ben.benson@raspberrypi.com>2024-06-06 11:15:08 +0100
committerKieran Bingham <kieran.bingham@ideasonboard.com>2024-06-13 11:02:16 +0100
commit8bea2d5a8adf1901f49e6449a731e3fd02272b3d (patch)
tree049daee7e8d6cf6ba49010dc14902b785754bee3 /utils/raspberrypi/ctt/ctt_dots_locator.py
parentd13542c28f291b88e29f1142d6dff565e772a05c (diff)
utils: raspberrypi: ctt: Added CAC support to the CTT
Added the ability to tune the chromatic aberration correction within the ctt. There are options for cac_only or to tune as part of a larger tuning process. CTT will now recognise any files that begin with "cac" as being chromatic aberration tuning files. Signed-off-by: Ben Benson <ben.benson@raspberrypi.com> Signed-off-by: David Plowman <david.plowman@raspberrypi.com> Reviewed-by: Naushir Patuck <naush@raspberrypi.com> Tested-by: Naushir Patuck <naush@raspberrypi.com> Acked-by: Kieran Bingham <kieran.bingham@ideasonboard.com> Signed-off-by: Kieran Bingham <kieran.bingham@ideasonboard.com>
Diffstat (limited to 'utils/raspberrypi/ctt/ctt_dots_locator.py')
-rw-r--r--utils/raspberrypi/ctt/ctt_dots_locator.py118
1 files changed, 118 insertions, 0 deletions
diff --git a/utils/raspberrypi/ctt/ctt_dots_locator.py b/utils/raspberrypi/ctt/ctt_dots_locator.py
new file mode 100644
index 00000000..4945c04b
--- /dev/null
+++ b/utils/raspberrypi/ctt/ctt_dots_locator.py
@@ -0,0 +1,118 @@
+# SPDX-License-Identifier: BSD-2-Clause
+#
+# Copyright (C) 2023, Raspberry Pi Ltd
+#
+# find_dots.py - Used by CAC algorithm to convert image to set of dots
+
+'''
+This file takes the black and white version of the image, along with
+the color version. It then located the black dots on the image by
+thresholding dark pixels.
+In a rather fun way, the algorithm bounces around the thresholded area in a random path
+We then use the maximum and minimum of these paths to determine the dot shape and size
+This info is then used to return colored dots and locations back to the main file
+'''
+
+import numpy as np
+import random
+from PIL import Image, ImageEnhance, ImageFilter
+
+
+def find_dots_locations(rgb_image, color_threshold=100, dots_edge_avoid=75, image_edge_avoid=10, search_path_length=500, grid_scan_step_size=10, logfile=open("log.txt", "a+")):
+ # Initialise some starting variables
+ pixels = Image.fromarray(rgb_image)
+ pixels = pixels.convert("L")
+ enhancer = ImageEnhance.Contrast(pixels)
+ im_output = enhancer.enhance(1.4)
+ # We smooth it slightly to make it easier for the dot recognition program to locate the dots
+ im_output = im_output.filter(ImageFilter.GaussianBlur(radius=2))
+ bw_image = np.array(im_output)
+
+ location = [0, 0]
+ dots = []
+ dots_location = []
+ # the program takes away the edges - we don't want a dot that is half a circle, the
+ # centroids would all be wrong
+ for x in range(dots_edge_avoid, len(bw_image) - dots_edge_avoid, grid_scan_step_size):
+ for y in range(dots_edge_avoid, len(bw_image[0]) - dots_edge_avoid, grid_scan_step_size):
+ location = [x, y]
+ scrap_dot = False # A variable used to make sure that this is a valid dot
+ if (bw_image[location[0], location[1]] < color_threshold) and not (scrap_dot):
+ heading = "south" # Define a starting direction to move in
+ coords = []
+ for i in range(search_path_length): # Creates a path of length `search_path_length`. This turns out to always be enough to work out the rough shape of the dot.
+ # Now make sure that the thresholded area doesn't come within 10 pixels of the edge of the image, ensures we capture all the CA
+ if ((image_edge_avoid < location[0] < len(bw_image) - image_edge_avoid) and (image_edge_avoid < location[1] < len(bw_image[0]) - image_edge_avoid)) and not (scrap_dot):
+ if heading == "south":
+ if bw_image[location[0] + 1, location[1]] < color_threshold:
+ # Here, notice it does not go south, but actually goes southeast
+ # This is crucial in ensuring that we make our way around the majority of the dot
+ location[0] = location[0] + 1
+ location[1] = location[1] + 1
+ heading = "south"
+ else:
+ # This happens when we reach a thresholded edge. We now randomly change direction and keep searching
+ dir = random.randint(1, 2)
+ if dir == 1:
+ heading = "west"
+ if dir == 2:
+ heading = "east"
+
+ if heading == "east":
+ if bw_image[location[0], location[1] + 1] < color_threshold:
+ location[1] = location[1] + 1
+ heading = "east"
+ else:
+ dir = random.randint(1, 2)
+ if dir == 1:
+ heading = "north"
+ if dir == 2:
+ heading = "south"
+
+ if heading == "west":
+ if bw_image[location[0], location[1] - 1] < color_threshold:
+ location[1] = location[1] - 1
+ heading = "west"
+ else:
+ dir = random.randint(1, 2)
+ if dir == 1:
+ heading = "north"
+ if dir == 2:
+ heading = "south"
+
+ if heading == "north":
+ if bw_image[location[0] - 1, location[1]] < color_threshold:
+ location[0] = location[0] - 1
+ heading = "north"
+ else:
+ dir = random.randint(1, 2)
+ if dir == 1:
+ heading = "west"
+ if dir == 2:
+ heading = "east"
+ # Log where our particle travels across the dot
+ coords.append([location[0], location[1]])
+ else:
+ scrap_dot = True # We just don't have enough space around the dot, discard this one, and move on
+ if not scrap_dot:
+ # get the size of the dot surrounding the dot
+ x_coords = np.array(coords)[:, 0]
+ y_coords = np.array(coords)[:, 1]
+ hsquaresize = max(list(x_coords)) - min(list(x_coords))
+ vsquaresize = max(list(y_coords)) - min(list(y_coords))
+ # Create the bounding coordinates of the rectangle surrounding the dot
+ # Program uses the dotsize + half of the dotsize to ensure we get all that color fringing
+ extra_space_factor = 0.45
+ top_left_x = (min(list(x_coords)) - int(hsquaresize * extra_space_factor))
+ btm_right_x = max(list(x_coords)) + int(hsquaresize * extra_space_factor)
+ top_left_y = (min(list(y_coords)) - int(vsquaresize * extra_space_factor))
+ btm_right_y = max(list(y_coords)) + int(vsquaresize * extra_space_factor)
+ # Overwrite the area of the dot to ensure we don't use it again
+ bw_image[top_left_x:btm_right_x, top_left_y:btm_right_y] = 255
+ # Add the color version of the dot to the list to send off, along with some coordinates.
+ dots.append(rgb_image[top_left_x:btm_right_x, top_left_y:btm_right_y])
+ dots_location.append([top_left_x, top_left_y])
+ else:
+ # Dot was too close to the image border to be useable
+ pass
+ return dots, dots_location