Commented Image Search

Author: Defcon27

FeatureVectors.py

@@ -5,12 +5,15 @@
 
 
 class FeatureVectors():
+    """Extracts various feature vectors from image """
 
     def __init__(self, image):
         self.filter = NoiseReduction(image)
         self.image = self.filter.applyGaussianBlur()
 
     def __getMeanIntensity(self):
+        # Returns the mean intensity of image
+
         meanIntensity = []
         for channel in range(3):
             channel_mean = np.average(self.image[:, :, channel])
@@ -19,6 +22,8 @@ def __getMeanIntensity(self):
         return meanIntensity
 
     def __getStdIntensity(self):
+        # Returns the standard deviation of intensity of image
+
         stdIntensity = []
         for channel in range(3):
             channel_std = np.std(self.image[:, :, channel])
@@ -27,6 +32,8 @@ def __getStdIntensity(self):
         return stdIntensity
 
     def __getRGBHistogramVector(self):
+        # Returns the 3-D(RGB) Histogram vector of image
+
         histogram_3d = cv2.calcHist([self.image], [0, 1, 2], None,
                                     [12, 12, 12], [0, 256, 0, 256, 0, 256])
         histogram_3d = histogram_3d.ravel()
@@ -35,13 +42,19 @@ def __getRGBHistogramVector(self):
         return RGBHistogram
 
     def __getHuMoments(self):
+        # Returns Hu-Moments vector of image
+
         filter = ConvolutionFilter(self.image)
         canny_filtered = filter.applyCannyEdge()
         canny_huMoments = cv2.HuMoments(cv2.moments(canny_filtered)).flatten()
         huVector = list(canny_huMoments.ravel())
         return huVector
 
     def getFeatureVector(self):
+        """ Return a python list of complete feature vectors
+        Extracts Statistics, 3-D Histogram, HuMoments from image and appends into single list
+        """
+
         featureVectors = []
         meanIntensity = self.__getMeanIntensity()
         stdIntensity = self.__getStdIntensity()

ImageSearch_Serial.py

@@ -7,6 +7,8 @@
 
 
 def extractFeatureVectors(image_path):
+    # Extracts feature vectors for input image
+
     image = cv2.imread(image_path)
     image = cv2.cvtColor(image, cv2.COLOR_BGR2RGB)
     image = cv2.resize(image, (500, 500))
@@ -18,6 +20,7 @@ def extractFeatureVectors(image_path):
 
 
 def ImageSearch(queryImage):
+    # Performs Image Search using Query image
 
     image_db_path = "Image_Database/"
     image_paths = []

ImageSearch_parallel.py

@@ -10,6 +10,7 @@
 
 
 def extractFeatureVectors(image_path):
+    # Extracts feature vectors for input image
 
     image = cv2.imread(image_path)
     image = cv2.cvtColor(image, cv2.COLOR_BGR2RGB)
@@ -22,6 +23,8 @@ def extractFeatureVectors(image_path):
 
 
 def ThreadedFeatureExtraction(images_list):
+     # Performing feature extraction of databse images using multithreading
+
     features = {}
     with concurrent.futures.ThreadPoolExecutor() as executor:
         results = [executor.submit(extractFeatureVectors, image_path)
@@ -42,6 +45,7 @@ def getImg(img):
 
 
 def ImageSearch(queryImage):
+    # Performs Image Search using Query image
 
     image_db_path = "Image_Database/"
     image_paths = []

QuerySearch.py

@@ -2,13 +2,16 @@
 
 
 class QuerySearch:
+    """ Searches similarity of query image vector with extracted vectors from database"""
 
     def __init__(self, queryFeature, features):
 
         self.features = features
         self.queryFeature = queryFeature
 
     def __cosine_similarity(self, queryVector, vector):
+        # Computes cosine similarity between two vectors
+
         eps = 1e-10
         dot_product = sum(q*v for q, v in zip(queryVector, vector))
 
@@ -23,6 +26,8 @@ def __cosine_similarity(self, queryVector, vector):
         return cosine_similarity
 
     def __chi2_distance(self, queryVector, vector):
+        # Computes chi-square distance between two vectors
+
         eps = 1e-10
         dists = [((q - v) ** 2) / (q + v + eps)
                  for (q, v) in zip(queryVector, vector)]
@@ -31,11 +36,15 @@ def __chi2_distance(self, queryVector, vector):
         return chi2_distance
 
     def __SAD_distance(self, queryVector, vector):
+        # Computes SAD distance between two vectors
+
         dists = [abs(q-v) for (q, v) in zip(queryVector, vector)]
         SAD_distance = sum(dists)
         return SAD_distance
 
     def performSearch(self):
+        # Return similarity indices of queryVector and databse images
+
         searchSimilarityScores = []
 
         for image in self.features: