Merge pull request #331 from kwcckw/dev

Updated PatternGenerator.

augraphy/augmentations/noisetexturize.py

@@ -7,7 +7,7 @@
 
 
 class NoiseTexturize(Augmentation):
-    """Creates a random noise based texture pattern to emulate paper textures.
+    """Creates a random noise pattern to emulate paper textures.
     Consequently applies noise patterns to the original image from big to small.
 
     :param sigma_range: Defines bounds of noise fluctuations.

augraphy/augmentations/quasicrystal.py

@@ -23,7 +23,6 @@
 
 """
 import math
-import os
 import random
 import warnings
 
@@ -39,34 +38,46 @@
 
 
 class PatternGenerator(Augmentation):
-    def __init__(self, imgx=512, imgy=512, n_rotation_range=(10, 15), p=1.0):
-        """
-        In this implementation we take a geometric plane and every point in the plane is shaded according
-        to its position,(x,y) coordinate. We take the pattern and perform a bitwise not operation so that it can
-        be added as an background to an image.This code is a python implementation of a QuasiPattern Distortion augmentation techniques
-        using PIL and the OpenCV libraries. This augmentation creates a new pattern image and superimposes it onto an input image.
-
-        To make the pattern more prominent
-        a. Increase the 'frequency' parameter: Increasing the frequency of the pattern will the it tightly populated and more prominent.
-        b. Decrease the 'n_rotation' parameter: Decreasing the number of rotations will make the pattern less symmetrical.
-
-        :param imgx: width of the pattern image. default is 512
-        :type imgx: int
-        :param imgy: height of the pattern image, default is 512
-        :type imgy: int
-        :param n_rotation: is the number of rotations applied to the pattern, default value lies
-                           between 10 and 15.
-        :type n_rotation: tuple (int)
-
-
-        """
+    """In this implementation we take a geometric plane and every point in the plane is shaded according
+    to its position,(x,y) coordinate. We take the pattern and perform a bitwise not operation so that it can
+    be added as an background to an image.This code is a python implementation of a QuasiPattern Distortion augmentation techniques
+    using PIL and the OpenCV libraries. This augmentation creates a new pattern image and superimposes it onto an input image.
+    To make the pattern more prominent
+    a. Increase the 'frequency' parameter: Increasing the frequency of the pattern will the it tightly populated and more prominent.
+    b. Decrease the 'n_rotation' parameter: Decreasing the number of rotations will make the pattern less symmetrical.
+
+    :param imgx: width of the pattern image. default is 512
+    :type imgx: int, optional
+    :param imgy: height of the pattern image, default is 512
+    :type imgy: int, optional
+    :param n_rotation: is the number of rotations applied to the pattern, default value lies
+                       between 10 and 15.
+    :type n_rotation: tuple (int) , optional
+    :param color: Color of the pattern in BGR format. Use "random" for random color effect.
+    :type color: tuple (int), optional
+    :param alpha_range: Tuple of floats determining the alpha value of the patterns.
+    :type alpha_range: tuple (float), optional
+    """
+
+    def __init__(
+        self,
+        imgx=512,
+        imgy=512,
+        n_rotation_range=(10, 15),
+        color="random",
+        alpha_range=(0.25, 0.5),
+        p=1.0,
+    ):
+        """Constructor method"""
         super().__init__(p=p)
         self.imgx = imgx  # width of the image
         self.imgy = imgy  # hieght of the image
         self.n_rotation_range = n_rotation_range  # number of rotation to be applied to the pattern
+        self.color = color
+        self.alpha_range = alpha_range
 
     def __repr__(self):
-        # return f"QuasiPattern Distortion: width = {self.imgx} , height = {self.imgy}, n_rotation = {self.n_rotation}"
+        # return f"QuasiPattern Distortion: width = {self.imgx} , height = {self.imgy}, n_rotation = {self.n_rotation}, color = {self.color}, alpha_range = {self.alpha_range}"
         return f"QuasiPattern Distortion:  n_rotation_range = {self.n_rotation_range}"
 
     @staticmethod
@@ -97,6 +108,7 @@ def __call__(self, image, layer=None, force=False):
         if force or self.should_run():
             result = image.copy()
             h, w = result.shape[:2]
+
             self.n_rotation = random.randint(self.n_rotation_range[0], self.n_rotation_range[1])
             pattern_image = np.zeros((self.imgy, self.imgx, 3), dtype=np.uint8)
             frequency = random.random() * 100 + 18  # determines the frequency of pattern
@@ -109,6 +121,19 @@ def __call__(self, image, layer=None, force=False):
                 invert = cv2.cvtColor(invert, cv2.COLOR_RGB2GRAY)
             elif len(image.shape) == 3 and image.shape[2] == 1:
                 invert = cv2.cvtColor(invert, cv2.COLOR_RGB2GRAY)
-            sw = PatternMaker(alpha=0.15)
+            sw = PatternMaker(alpha=random.uniform(self.alpha_range[0], self.alpha_range[1]))
+
+            # apply color into pattern
+            if self.color == "random":
+                color = (random.randint(0, 255), random.randint(0, 255), random.randint(0, 255))
+            else:
+                color = self.color
+            if len(invert.shape) > 2:
+                color_mask = np.full_like(invert, fill_value=color, dtype="uint8")
+            else:
+                color_mask = np.full_like(invert, fill_value=np.mean(color), dtype="uint8")
+            invert = cv2.multiply(invert, color_mask, scale=1 / 255)
+
+            # overlay pattern into image
             result = sw.superimpose(result, invert)
             return result

doc/source/augmentations/noisetexturize.rst

@@ -10,7 +10,7 @@ NoiseTexturize
 --------
 Overview
 --------
-The Noise Texturize augmentation creates a random noise based texture pattern to emulate paper textures.
+The Noise Texturize augmentation creates a random noise pattern to emulate paper textures.
 
 Initially, a clean image with single line of text is created.
 
@@ -45,15 +45,15 @@ Clean image:
 ---------
 Example 1
 ---------
-In this example, a NoiseTexturize augmentation instance is initialized and the sigma value that define noise fluctuatiosn is set in between 12 and 15 (12, 15).
-The noise turbulence range is set in between 3 and 5.
+In this example, a NoiseTexturize augmentation instance is initialized and the sigma value that define noise fluctuatiosn is set in between 3 and 10 (3, 10).
+The noise turbulence range is set in between 2 and 5.
 
 Code example:
 
 ::
 
-    noise_texturize = NoiseTexturize(sigma_range=(12, 15),
-                                     turbulence_range=(3, 5),
+    noise_texturize = NoiseTexturize(sigma_range=(3, 10),
+                                     turbulence_range=(2, 5),
                                      )
 
     img_noise_texturize = noise_texturize(image)