first commit

.gitignore

@@ -0,0 +1 @@
+".idea" 

README.md

@@ -0,0 +1 @@
+"# UtilityMeterNumberRecognition" 

dao/__init__.py

@@ -0,0 +1 @@
+__author__ = 'Krtalici'

dao/trainingDao.py

@@ -0,0 +1,37 @@
+__author__ = 'Krtalici'
+import pickle as pkl
+import numpy as np
+
+def writeTrainingData( imagesInVectorForm ) :
+    np.savetxt('test.txt', imagesInVectorForm)
+
+def writeTargets( ) :
+    outcomes = open('trainingData/target.pkl', 'w')
+
+    #target = np.array([0,0,0,1,0,0,1,0,0,0,1,0,1,0,1,0,0,0,0,0])
+    target = np.array([0,1,0,0,1])
+
+    pkl.dump(target, outcomes)
+    outcomes.close()
+
+def readTrainingData() :
+    imagesInVectorForm = np.loadtxt('test.txt')
+    imagesInVectorForm = np.delete(imagesInVectorForm, 0, 0)
+    print "Size [%d,%d]"%imagesInVectorForm.shape
+
+    #matrica trening slika, svaka slika pretvorena pri zapisivanju u datoteku u vektor
+    return imagesInVectorForm
+
+def readTrainingTarget() :
+    pkl_file = open('trainingData/target.pkl', 'r')
+    target = pkl.load(pkl_file)
+    pkl_file.close()
+    #matrica trening slika, svaka slika pretvorena pri zapisivanju u datoteku u vektor
+    return target
+
+def pickleLoader(pklFile):
+    try:
+        while True:
+            yield pkl.load(pklFile)
+    except EOFError:
+        pass

display/__init__.py

@@ -0,0 +1 @@
+__author__ = 'Krtalici'

display/display.py

@@ -0,0 +1,9 @@
+__author__ = 'Krtalici'
+import cv2
+
+def display_images(images):
+    for image in images:
+        cv2.imshow("Image", image)
+        cv2.waitKey(0)
+
+

image_preprocesing/__init__.py

@@ -0,0 +1 @@
+__author__ = 'Krtalici'

image_preprocesing/contrast_streching.py

@@ -0,0 +1,19 @@
+import numpy as np
+import cv2
+from skimage import data, img_as_float
+from skimage import exposure
+from image_preprocesing import image_enhancment as ie
+
+
+if __name__ == "__main__":
+    image = cv2.imread('../watermeter21.png', cv2.IMREAD_GRAYSCALE)
+    cv2.imshow('orig', image)
+    p2, p98 = np.percentile(image, (2, 97))
+    img_rescale = exposure.rescale_intensity(image, in_range=(p2, p98))
+    image_to_return = ie.sharp_img_with_filter(img_rescale)
+    final = cv2.bilateralFilter(image_to_return, 9, 75, 75)
+    p2, p98 = np.percentile(image, (2, 97))
+    img_rescale = exposure.rescale_intensity(final, in_range=(p2, p98))
+    cv2.imshow('Streched', img_rescale)
+    cv2.imwrite('../watermeter23.png', img_rescale)
+    cv2.waitKey(0)

image_preprocesing/filter_countours.py

@@ -0,0 +1,14 @@
+__author__ = 'Krtalici'
+import cv2
+
+def filterByHierarchy(cnts,hierarchy):
+    cntsWithMAxHierarchy = []
+    for i,cnt in enumerate(cnts):
+        if hierarchy[0,i,3] == -1  and cv2.contourArea(cnt[0]) > 80:
+            cntsWithMAxHierarchy.append(cnt)
+
+    return cntsWithMAxHierarchy
+
+
+
+

image_preprocesing/histogram.py

@@ -0,0 +1,83 @@
+__author__ = 'Krtalici'
+import cv2
+import numpy as np
+from image_preprocesing import image_tresholding as tresh
+from image_preprocesing import image_enhancment as ie
+
+def get_grey_hist_eq( image ):
+    image = cv2.cvtColor(image, cv2.COLOR_BGR2GRAY)
+    eq = cv2.equalizeHist(image)
+    return eq
+
+def sharp_img(path_to_image):
+
+    #Load source / input image as grayscale, also works on color images...
+    imgIn = cv2.imread(path_to_image, cv2.IMREAD_GRAYSCALE)
+    cv2.imshow("Original", imgIn)
+    blur = cv2.bilateralFilter(imgIn, 9, 75, 75)
+    #blur = cv2.medianBlur(imgIn, 7)
+
+    #Create the identity filter, but with the 1 shifted to the right!
+    kernel = np.zeros((9, 9), np.float32)
+    kernel[4, 4] = 2.0   #Identity, times two!
+
+    #Create a box filter:
+    boxFilter = np.ones((9, 9), np.float32) / 81.0
+
+    #Subtract the two:
+    kernel = kernel - boxFilter
+
+
+    #Note that we are subject to overflow and underflow here...but I believe that
+    # filter2D clips top and bottom ranges on the output, plus you'd need a
+    # very bright or very dark pixel surrounded by the opposite type.
+
+    custom = cv2.filter2D(blur, -1, kernel)
+    cv2.imshow("Sharpen", custom)
+
+    customSharp = cv2.filter2D(custom, -1, kernel)
+    cv2.imshow("Custom Sharpen", customSharp)
+
+    treshed = tresh.treshImageOtsuWithCorrection(customSharp, 0)
+    cv2.imshow('Treshed', treshed)
+    cv2.waitKey(0)
+
+def sharp_that_image(grey_image):
+
+    #Load source / input image as grayscale, also works on color images...
+    cv2.imshow("Original", grey_image)
+    blur = cv2.bilateralFilter(grey_image, 9, 75, 75)
+    #blur = cv2.medianBlur(imgIn, 7)
+
+    #Create the identity filter, but with the 1 shifted to the right!
+    kernel = np.zeros((9, 9), np.float32)
+    kernel[4, 4] = 2.0   #Identity, times two!
+
+    #Create a box filter:
+    boxFilter = np.ones((9, 9), np.float32) / 81.0
+
+    #Subtract the two:
+    kernel = kernel - boxFilter
+
+
+    #Note that we are subject to overflow and underflow here...but I believe that
+    # filter2D clips top and bottom ranges on the output, plus you'd need a
+    # very bright or very dark pixel surrounded by the opposite type.
+
+    custom = cv2.filter2D(blur, -1, kernel)
+    cv2.imshow("Sharpen", custom)
+
+    #customSharp = cv2.filter2D(custom, -1, kernel)
+    #cv2.imshow("Custom Sharpen", customSharp)
+
+    treshed = tresh.treshImageOtsuWithCorrection(custom, 5)
+    cv2.imshow('Treshed', treshed)
+    cv2.waitKey(0)
+
+if __name__ == "__main__":
+    image = cv2.imread('../watermeter21.png')
+    eq = get_grey_hist_eq(image)
+    image = cv2.cvtColor(image, cv2.COLOR_BGR2GRAY)
+    cv2.imshow("Histogram Equalization", np.hstack([image, eq]))
+    cv2.waitKey(0)
+    sharp_that_image(grey_image=eq)