extract.py

import numpy as np
import cv2
import math
# from matplotlib import pyplot as plt

# please don't worry about these variables now
ANCHOR_POINT = 6000
MIDZONE_THRESHOLD = 15000
MIN_HANDWRITING_HEIGHT_PIXEL = 20

# Features are defined here as global variables
BASELINE_ANGLE = 0.0
TOP_MARGIN = 0.0
LETTER_SIZE = 0.0
LINE_SPACING = 0.0
WORD_SPACING = 0.0
PEN_PRESSURE = 0.0
SLANT_ANGLE = 0.0

''' function for bilateral filtering '''


def bilateralFilter(image, d):
    image = cv2.bilateralFilter(image, d, 50, 50)
    return image


''' function for median filtering '''


def medianFilter(image, d):
    image = cv2.medianBlur(image, d)
    return image


''' function for INVERTED binary threshold '''


def threshold(image, t):
    image = cv2.cvtColor(image, cv2.COLOR_BGR2GRAY)
    ret, image = cv2.threshold(image, t, 255, cv2.THRESH_BINARY_INV)
    return image


''' function for dilation of objects in the image '''


def dilate(image, kernalSize):
    kernel = np.ones(kernalSize, np.uint8)
    image = cv2.dilate(image, kernel, iterations=1)
    return image


''' function for erosion of objects in the image '''


def erode(image, kernalSize):
    kernel = np.ones(kernalSize, np.uint8)
    image = cv2.erode(image, kernel, iterations=1)
    return image


''' function for finding contours and straightening them horizontally. Straightened lines will give better result with horizontal projections. '''


def straighten(image):

    global BASELINE_ANGLE

    angle = 0.0
    angle_sum = 0.0
    contour_count = 0

    # these four variables are not being used, please ignore
    positive_angle_sum = 0.0  # downward
    negative_angle_sum = 0.0  # upward
    positive_count = 0
    negative_count = 0

    # apply bilateral filter
    filtered = bilateralFilter(image, 3)
    # cv2.imshow('filtered',filtered)

    # convert to grayscale and binarize the image by INVERTED binary thresholding
    thresh = threshold(filtered, 120)
    # cv2.imshow('thresh',thresh)

    # dilate the handwritten lines in image with a suitable kernel for contour operation
    dilated = dilate(thresh, (5, 100))
    # cv2.imshow('dilated',dilated)

    im2, ctrs, hier = cv2.findContours(
        dilated.copy(), cv2.RETR_EXTERNAL, cv2.CHAIN_APPROX_SIMPLE)

    for i, ctr in enumerate(ctrs):
        x, y, w, h = cv2.boundingRect(ctr)

        # We can be sure the contour is not a line if height > width or height is < 20 pixels. Here 20 is arbitrary.
        if h > w or h < MIN_HANDWRITING_HEIGHT_PIXEL:
            continue

        # We extract the region of interest/contour to be straightened.
        roi = image[y:y+h, x:x+w]
        # rows, cols = ctr.shape[:2]

        # If the length of the line is less than one third of the document width, especially for the last line,
        # ignore because it may yeild inacurate baseline angle which subsequently affects proceeding features.

        if w < image.shape[1]/2:
            roi = 255
            image[y:y+h, x:x+w] = roi
            continue

        # minAreaRect is necessary for straightening
        rect = cv2.minAreaRect(ctr)
        center = rect[0]
        angle = rect[2]
        # print "original: "+str(i)+" "+str(angle)
        # I actually gave a thought to this but hard to remember anyway!
        if angle < -45.0:
            angle += 90.0
        # print "+90 "+str(i)+" "+str(angle)

        rot = cv2.getRotationMatrix2D(((x+w)/2, (y+h)/2), angle, 1)
        # extract = cv2.warpAffine(roi, rot, (w,h), borderMode=cv2.BORDER_TRANSPARENT)
        extract = cv2.warpAffine(
            roi, rot, (w, h), borderMode=cv2.BORDER_CONSTANT, borderValue=(255, 255, 255))
        # cv2.imshow('warpAffine:'+str(i),extract)

        # image is overwritten with the straightened contour
        image[y:y+h, x:x+w] = extract
        '''
		# Please Ignore. This is to draw visual representation of the contour rotation.
		box = cv2.boxPoints(rect)
		box = np.int0(box)
		cv2.drawContours(display,[box],0,(0,0,255),1)
		cv2.rectangle(display,(x,y),( x + w, y + h ),(0,255,0),1)
		'''
        # print angle
        angle_sum += angle
        coutour_count += 1
    '''	
		# sum of all the angles of downward baseline
		if(angle>0.0):
			positive_angle_sum += angle
			positive_count += 1
		# sum of all the angles of upward baseline
		else:
			negative_angle_sum += angle
			negative_count += 1
			
	if(positive_count == 0): positive_count = 1
	if(negative_count == 0): negative_count = 1
	average_positive_angle = positive_angle_sum / positive_count
	average_negative_angle = negative_angle_sum / negative_count
	print "average_positive_angle: "+str(average_positive_angle)
	print "average_negative_angle: "+str(average_negative_angle)
	
	if(abs(average_positive_angle) > abs(average_negative_angle)):
		average_angle = average_positive_angle
	else:
		average_angle = average_negative_angle
	
	print "average_angle: "+str(average_angle)
	'''
    # cv2.imshow('contours', display)

    # mean angle of the contours (not lines) is found
    mean_angle = angle_sum / contour_count
    BASELINE_ANGLE = mean_angle
    # print ("Average baseline angle: "+str(mean_angle))
    return image


''' function to calculate horizontal projection of the image pixel rows and return it '''


def horizontalProjection(img):
    # Return a list containing the sum of the pixels in each row
    (h, w) = img.shape[:2]
    sumRows = []
    for j in range(h):
        row = img[j:j+1, 0:w]  # y1:y2, x1:x2
        sumRows.append(np.sum(row))
    return sumRows


''' function to calculate vertical projection of the image pixel columns and return it '''


def verticalProjection(img):
    # Return a list containing the sum of the pixels in each column
    (h, w) = img.shape[:2]
    sumCols = []
    for j in range(w):
        col = img[0:h, j:j+1]  # y1:y2, x1:x2
        sumCols.append(np.sum(col))
    return sumCols


''' function to extract lines of handwritten text from the image using horizontal projection '''


def extractLines(img):

    global LETTER_SIZE
    global LINE_SPACING
    global TOP_MARGIN

    # apply bilateral filter
    filtered = bilateralFilter(img, 5)

    # convert to grayscale and binarize the image by INVERTED binary thresholding
    # it's better to clear unwanted dark areas at the document left edge and use a high threshold value to preserve more text pixels
    thresh = threshold(filtered, 160)
    # cv2.imshow('thresh', lthresh)

    # extract a python list containing values of the horizontal projection of the image into 'hp'
    hpList = horizontalProjection(thresh)

    # Extracting 'Top Margin' feature.
    topMarginCount = 0
    for sum in hpList:
        # sum can be strictly 0 as well. Anyway we take 0 and 255.
        if (sum <= 255):
            topMarginCount += 1
        else:
            break

    # print "(Top margin row count: "+str(topMarginCount)+")"

    # FIRST we extract the straightened contours from the image by looking at occurance of 0's in the horizontal projection.
    lineTop = 0
    lineBottom = 0
    spaceTop = 0
    spaceBottom = 0
    indexCount = 0
    setLineTop = True
    setSpaceTop = True
    includeNextSpace = True
    space_zero = []  # stores the amount of space between lines
    lines = []  # a 2D list storing the vertical start index and end index of each contour

    # we are scanning the whole horizontal projection now
    for i, sum in enumerate(hpList):
        # sum being 0 means blank space
        if (sum == 0):
            if (setSpaceTop):
                spaceTop = indexCount
                setSpaceTop = False  # spaceTop will be set once for each start of a space between lines
            indexCount += 1
            spaceBottom = indexCount
            if (i < len(hpList)-1):  # this condition is necessary to avoid array index out of bound error
                # if the next horizontal projectin is 0, keep on counting, it's still in blank space
                if (hpList[i+1] == 0):
                    continue
            # we are using this condition if the previous contour is very thin and possibly not a line
            if (includeNextSpace):
                space_zero.append(spaceBottom-spaceTop)
            else:
                if (len(space_zero) == 0):
                    previous = 0
                else:
                    previous = space_zero.pop()
                space_zero.append(previous + spaceBottom-lineTop)
            # next time we encounter 0, it's begining of another space so we set new spaceTop
            setSpaceTop = True

        # sum greater than 0 means contour
        if (sum > 0):
            if (setLineTop):
                lineTop = indexCount
                setLineTop = False  # lineTop will be set once for each start of a new line/contour
            indexCount += 1
            lineBottom = indexCount
            if (i < len(hpList)-1):  # this condition is necessary to avoid array index out of bound error
                # if the next horizontal projectin is > 0, keep on counting, it's still in contour
                if (hpList[i+1] > 0):
                    continue

                # if the line/contour is too thin <10 pixels (arbitrary) in height, we ignore it.
                # Also, we add the space following this and this contour itself to the previous space to form a bigger space: spaceBottom-lineTop.
                if (lineBottom-lineTop < 20):
                    includeNextSpace = False
                    setLineTop = True  # next time we encounter value > 0, it's begining of another line/contour so we set new lineTop
                    continue
            # the line/contour is accepted, new space following it will be accepted
            includeNextSpace = True

            # append the top and bottom horizontal indices of the line/contour in 'lines'
            lines.append([lineTop, lineBottom])
            setLineTop = True  # next time we encounter value > 0, it's begining of another line/contour so we set new lineTop

    '''
	# Printing the values we found so far.
	for i, line in enumerate(lines):
		print
		print i
		print line[0]
		print line[1]
		print len(hpList[line[0]:line[1]])
		print hpList[line[0]:line[1]]
	
	for i, line in enumerate(lines):
		cv2.imshow("line "+str(i), img[line[0]:line[1], : ])
	'''

    # SECOND we extract the very individual lines from the lines/contours we extracted above.
    fineLines = []  # a 2D list storing the horizontal start index and end index of each individual line
    for i, line in enumerate(lines):

        # 'anchor' will locate the horizontal indices where horizontal projection is > ANCHOR_POINT for uphill or < ANCHOR_POINT for downhill(ANCHOR_POINT is arbitrary yet suitable!)
        anchor = line[0]
        anchorPoints = []  # python list where the indices obtained by 'anchor' will be stored
        # it implies that we expect to find the start of an individual line (vertically), climbing up the histogram
        upHill = True
        # it implies that we expect to find the end of an individual line (vertically), climbing down the histogram
        downHill = False
        # we put the region of interest of the horizontal projection of each contour here
        segment = hpList[line[0]:line[1]]

        for j, sum in enumerate(segment):
            if (upHill):
                if (sum < ANCHOR_POINT):
                    anchor += 1
                    continue
                anchorPoints.append(anchor)
                upHill = False
                downHill = True
            if (downHill):
                if (sum > ANCHOR_POINT):
                    anchor += 1
                    continue
                anchorPoints.append(anchor)
                downHill = False
                upHill = True

        # print anchorPoints

        # we can ignore the contour here
        if (len(anchorPoints) < 2):
            continue

        '''
		# the contour turns out to be an individual line
		if(len(anchorPoints)<=3):
			fineLines.append(line)
			continue
		'''
        # len(anchorPoints) > 3 meaning contour composed of multiple lines
        lineTop = line[0]
        for x in range(1, len(anchorPoints)-1, 2):
            # 'lineMid' is the horizontal index where the segmentation will be done
            lineMid = (anchorPoints[x]+anchorPoints[x+1])/2
            lineBottom = lineMid
            # line having height of pixels <20 is considered defects, so we just ignore it
            # this is a weakness of the algorithm to extract lines (anchor value is ANCHOR_POINT, see for different values!)
            if (lineBottom-lineTop < 20):
                continue
            fineLines.append([lineTop, lineBottom])
            lineTop = lineBottom
        if (line[1]-lineTop < 20):
            continue
        fineLines.append([lineTop, line[1]])

    # LINE SPACING and LETTER SIZE will be extracted here
    # We will count the total number of pixel rows containing upper and lower zones of the lines and add the space_zero/runs of 0's(excluding first and last of the list ) to it.
    # We will count the total number of pixel rows containing midzones of the lines for letter size.
    # For this, we set an arbitrary (yet suitable!) threshold MIDZONE_THRESHOLD = 15000 in horizontal projection to identify the midzone containing rows.
    # These two total numbers will be divided by number of lines (having at least one row>MIDZONE_THRESHOLD) to find average line spacing and average letter size.
    space_nonzero_row_count = 0
    midzone_row_count = 0
    lines_having_midzone_count = 0
    flag = False
    for i, line in enumerate(fineLines):
        segment = hpList[line[0]:line[1]]
        for j, sum in enumerate(segment):
            if (sum < MIDZONE_THRESHOLD):
                space_nonzero_row_count += 1
            else:
                midzone_row_count += 1
                flag = True

        # This line has contributed at least one count of pixel row of midzone
        if (flag):
            lines_having_midzone_count += 1
            flag = False

    # error prevention ^-^
    if (lines_having_midzone_count == 0):
        lines_having_midzone_count = 1

    # excluding first and last entries: Top and Bottom margins
    total_space_row_count = space_nonzero_row_count + np.sum(space_zero[1:-1])
    # the number of spaces is 1 less than number of lines but total_space_row_count contains the top and bottom spaces of the line
    average_line_spacing = float(
        total_space_row_count) / lines_having_midzone_count
    average_letter_size = float(midzone_row_count) / lines_having_midzone_count
    # letter size is actually height of the letter and we are not considering width
    LETTER_SIZE = average_letter_size
    # error prevention ^-^
    if (average_letter_size == 0):
        average_letter_size = 1
    # We can't just take the average_line_spacing as a feature directly. We must take the average_line_spacing relative to average_letter_size.
    # Let's take the ratio of average_line_spacing to average_letter_size as the LINE SPACING, which is perspective to average_letter_size.
    relative_line_spacing = average_line_spacing / average_letter_size
    LINE_SPACING = relative_line_spacing

    # Top marging is also taken relative to average letter size of the handwritting
    relative_top_margin = float(topMarginCount) / average_letter_size
    TOP_MARGIN = relative_top_margin

    '''
	# showing the final extracted lines
	for i, line in enumerate(fineLines):
		cv2.imshow("line "+str(i), img[line[0]:line[1], : ])
	'''

    # print space_zero
    # print lines
    # print fineLines
    # print midzone_row_count
    # print total_space_row_count
    # print len(hpList)
    # print average_line_spacing
    # print lines_having_midzone_count
    # print i
    '''
	print ("Average letter size: "+str(average_letter_size))
	print ("Top margin relative to average letter size: "+str(relative_top_margin))
	print ("Average line spacing relative to average letter size: "+str(relative_line_spacing))
	'''
    return fineLines


''' function to extract words from the lines using vertical projection '''


def extractWords(image, lines):

    global LETTER_SIZE
    global WORD_SPACING

    # apply bilateral filter
    filtered = bilateralFilter(image, 5)

    # convert to grayscale and binarize the image by INVERTED binary thresholding
    thresh = threshold(filtered, 180)
    # cv2.imshow('thresh', wthresh)

    # Width of the whole document is found once.
    width = thresh.shape[1]
    space_zero = []  # stores the amount of space between words
    words = []  # a 2D list storing the coordinates of each word: y1, y2, x1, x2

    # Isolated words or components will be extacted from each line by looking at occurance of 0's in its vertical projection.
    for i, line in enumerate(lines):
        extract = thresh[line[0]:line[1], 0:width]  # y1:y2, x1:x2
        vp = verticalProjection(extract)
        # print i
        # print vp

        wordStart = 0
        wordEnd = 0
        spaceStart = 0
        spaceEnd = 0
        indexCount = 0
        setWordStart = True
        setSpaceStart = True
        includeNextSpace = True
        spaces = []

        # we are scanning the vertical projection
        for j, sum in enumerate(vp):
            # sum being 0 means blank space
            if (sum == 0):
                if (setSpaceStart):
                    spaceStart = indexCount
                    # spaceStart will be set once for each start of a space between lines
                    setSpaceStart = False
                indexCount += 1
                spaceEnd = indexCount
                if (j < len(vp)-1):  # this condition is necessary to avoid array index out of bound error
                    # if the next vertical projectin is 0, keep on counting, it's still in blank space
                    if (vp[j+1] == 0):
                        continue

                # we ignore spaces which is smaller than half the average letter size
                if ((spaceEnd-spaceStart) > int(LETTER_SIZE/2)):
                    spaces.append(spaceEnd-spaceStart)

                # next time we encounter 0, it's begining of another space so we set new spaceStart
                setSpaceStart = True

            # sum greater than 0 means word/component
            if (sum > 0):
                if (setWordStart):
                    wordStart = indexCount
                    setWordStart = False  # wordStart will be set once for each start of a new word/component
                indexCount += 1
                wordEnd = indexCount
                if (j < len(vp)-1):  # this condition is necessary to avoid array index out of bound error
                    # if the next horizontal projectin is > 0, keep on counting, it's still in non-space zone
                    if (vp[j+1] > 0):
                        continue

                # append the coordinates of each word/component: y1, y2, x1, x2 in 'words'
                # we ignore the ones which has height smaller than half the average letter size
                # this will remove full stops and commas as an individual component
                count = 0
                for k in range(line[1]-line[0]):
                    row = thresh[line[0]+k:line[0]+k+1,
                                 wordStart:wordEnd]  # y1:y2, x1:x2
                    if (np.sum(row)):
                        count += 1
                if (count > int(LETTER_SIZE/2)):
                    words.append([line[0], line[1], wordStart, wordEnd])

                # next time we encounter value > 0, it's begining of another word/component so we set new wordStart
                setWordStart = True

        space_zero.extend(spaces[1:-1])

    # print space_zero
    space_columns = np.sum(space_zero)
    space_count = len(space_zero)
    if (space_count == 0):
        space_count = 1
    average_word_spacing = float(space_columns) / space_count
    relative_word_spacing = average_word_spacing / LETTER_SIZE
    WORD_SPACING = relative_word_spacing
    # print "Average word spacing: "+str(average_word_spacing)
    # print ("Average word spacing relative to average letter size: "+str(relative_word_spacing))

    return words


''' function to determine the average slant of the handwriting '''


def extractSlant(img, words):

    global SLANT_ANGLE
    '''
	0.01 radian = 0.5729578 degree :: I had to put this instead of 0.0 becuase there was a bug yeilding inacurate value which I could not figure out!
	5 degree = 0.0872665 radian :: Hardly noticeable or a very little slant
	15 degree = 0.261799 radian :: Easily noticeable or average slant
	30 degree = 0.523599 radian :: Above average slant
	45 degree = 0.785398 radian :: Extreme slant
	'''
    # We are checking for 9 different values of angle
    theta = [-0.785398, -0.523599, -0.261799, -0.0872665,
             0.01, 0.0872665, 0.261799, 0.523599, 0.785398]
    # theta = [-0.785398, -0.523599, -0.436332, -0.349066, -0.261799, -0.174533, -0.0872665, 0, 0.0872665, 0.174533, 0.261799, 0.349066, 0.436332, 0.523599, 0.785398]

    # Corresponding index of the biggest value in s_function will be the index of the most likely angle in 'theta'
    s_function = [0.0] * 9
    count_ = [0]*9

    # apply bilateral filter
    filtered = bilateralFilter(img, 5)

    # convert to grayscale and binarize the image by INVERTED binary thresholding
    # it's better to clear unwanted dark areas at the document left edge and use a high threshold value to preserve more text pixels
    thresh = threshold(filtered, 180)
    # cv2.imshow('thresh', lthresh)

    # loop for each value of angle in theta
    for i, angle in enumerate(theta):
        s_temp = 0.0  # overall sum of the functions of all the columns of all the words!
        count = 0  # just counting the number of columns considered to contain a vertical stroke and thus contributing to s_temp

        # loop for each word
        for j, word in enumerate(words):
            original = thresh[word[0]:word[1], word[2]:word[3]]  # y1:y2, x1:x2

            height = word[1]-word[0]
            width = word[3]-word[2]

            # the distance in pixel we will shift for affine transformation
            # it's divided by 2 because the uppermost point and the lowermost points are being equally shifted in opposite directions
            shift = (math.tan(angle) * height) / 2

            # the amount of extra space we need to add to the original image to preserve information
            # yes, this is adding more number of columns but the effect of this will be negligible
            pad_length = abs(int(shift))

            # create a new image that can perfectly hold the transformed and thus widened image
            blank_image = np.zeros((height, width+pad_length*2, 3), np.uint8)
            new_image = cv2.cvtColor(blank_image, cv2.COLOR_BGR2GRAY)
            new_image[:, pad_length:width+pad_length] = original

            # points to consider for affine transformation
            (height, width) = new_image.shape[:2]
            x1 = width/2
            y1 = 0
            x2 = width/4
            y2 = height
            x3 = 3*width/4
            y3 = height

            pts1 = np.float32([[x1, y1], [x2, y2], [x3, y3]])
            pts2 = np.float32([[x1+shift, y1], [x2-shift, y2], [x3-shift, y3]])
            M = cv2.getAffineTransform(pts1, pts2)
            deslanted = cv2.warpAffine(new_image, M, (width, height))

            # find the vertical projection on the transformed image
            vp = verticalProjection(deslanted)

            # loop for each value of vertical projection, which is for each column in the word image
            for k, sum in enumerate(vp):
                # the columns is empty
                if (sum == 0):
                    continue

                # this is the number of foreground pixels in the column being considered
                num_fgpixel = sum / 255

                # if number of foreground pixels is less than onethird of total pixels, it is not a vertical stroke so we can ignore
                if (num_fgpixel < int(height/3)):
                    continue

                # the column itself is extracted, and flattened for easy operation
                column = deslanted[0:height, k:k+1]
                column = column.flatten()

                # now we are going to find the distance between topmost pixel and bottom-most pixel
                # l counts the number of empty pixels from top until and upto a foreground pixel is discovered
                for l, pixel in enumerate(column):
                    if (pixel == 0):
                        continue
                    break
                # m counts the number of empty pixels from bottom until and upto a foreground pixel is discovered
                for m, pixel in enumerate(column[::-1]):
                    if (pixel == 0):
                        continue
                    break

                # the distance is found as delta_y, I just followed the naming convention in the research paper I followed
                delta_y = height - (l+m)

                # please refer the research paper for more details of this function, anyway it's nothing tricky
                h_sq = (float(num_fgpixel)/delta_y)**2

                # I am multiplying by a factor of num_fgpixel/height to the above function to yeild better result
                # this will also somewhat negate the effect of adding more columns and different column counts in the transformed image of the same word
                h_wted = (h_sq * num_fgpixel) / height

                '''
				# just printing
				if(j==0):
					print column
					print str(i)+' h_sq='+str(h_sq)+' h_wted='+str(h_wted)+' num_fgpixel='+str(num_fgpixel)+' delta_y='+str(delta_y)
				'''

                # add up the values from all the loops of ALL the columns of ALL the words in the image
                s_temp += h_wted

                count += 1

            '''
			if(j==0):
				#plt.subplot(),plt.imshow(deslanted),plt.title('Output '+str(i))
				#plt.show()
				cv2.imshow('Output '+str(i)+str(j), deslanted)
				#print vp
				#print 'line '+str(i)+' '+str(s_temp)
				#print
			'''

        s_function[i] = s_temp
        count_[i] = count

    # finding the largest value and corresponding index
    max_value = 0.0
    max_index = 4
    for index, value in enumerate(s_function):
        # print str(index)+" "+str(value)+" "+str(count_[index])
        if (value > max_value):
            max_value = value
            max_index = index

    # We will add another value 9 manually to indicate irregular slant behaviour.
    # This will be seen as value 4 (no slant) but 2 corresponding angles of opposite sign will have very close values.
    if (max_index == 0):
        angle = 45
        result = " : Extremely right slanted"
    elif (max_index == 1):
        angle = 30
        result = " : Above average right slanted"
    elif (max_index == 2):
        angle = 15
        result = " : Average right slanted"
    elif (max_index == 3):
        angle = 5
        result = " : A little right slanted"
    elif (max_index == 5):
        angle = -5
        result = " : A little left slanted"
    elif (max_index == 6):
        angle = -15
        result = " : Average left slanted"
    elif (max_index == 7):
        angle = -30
        result = " : Above average left slanted"
    elif (max_index == 8):
        angle = -45
        result = " : Extremely left slanted"
    elif (max_index == 4):
        p = s_function[4] / s_function[3]
        q = s_function[4] / s_function[5]
        # print 'p='+str(p)+' q='+str(q)
        # the constants here are abritrary but I think suits the best
        if ((p <= 1.2 and q <= 1.2) or (p > 1.4 and q > 1.4)):
            angle = 0
            result = " : No slant"
        elif ((p <= 1.2 and q-p > 0.4) or (q <= 1.2 and p-q > 0.4)):
            angle = 0
            result = " : No slant"
        else:
            max_index = 9
            angle = 180
            result = " : Irregular slant behaviour"

        if angle == 0:
            print "\n************************************************"
            print "Slant determined to be straight."
        else:
            print "\n************************************************"
            print "Slant determined to be irregular."
        cv2.imshow("Check Image", img)
        cv2.waitKey(0)
        cv2.destroyAllWindows()
        type = raw_input("Press enter if okay, else enter c to change: ")
        if type == 'c':
            if angle == 0:
                angle = 180
                result = " : Irregular Slant"
                print "Set as"+result
                print "************************************************\n"
            else:
                angle = 0
                result = " : Straight/No Slant"
                print "Set as"+result
                print "************************************************\n"
        else:
            print "No Change!"
            print "************************************************\n"

    SLANT_ANGLE = angle
    # print ("Slant angle(degree): "+str(SLANT_ANGLE)+result)
    return


''' function to extract average pen pressure of the handwriting '''


def barometer(image):

    global PEN_PRESSURE

    # it's extremely necessary to convert to grayscale first
    image = cv2.cvtColor(image, cv2.COLOR_BGR2GRAY)
    # inverting the image pixel by pixel individually. This costs the maximum time and processing in the entire process!
    h, w = image.shape[:]
    inverted = image
    for x in range(h):
        for y in range(w):
            inverted[x][y] = 255 - image[x][y]

    # cv2.imshow('inverted', inverted)

    # bilateral filtering
    filtered = bilateralFilter(inverted, 3)

    # binary thresholding. Here we use 'threshold to zero' which is crucial for what we want.
    # If src(x,y) is lower than threshold=100, the new pixel value will be set to 0, else it will be left untouched!
    ret, thresh = cv2.threshold(filtered, 100, 255, cv2.THRESH_TOZERO)
    # cv2.imshow('thresh', thresh)

    # add up all the non-zero pixel values in the image and divide by the number of them to find the average pixel value in the whole image
    total_intensity = 0
    pixel_count = 0
    for x in range(h):
        for y in range(w):
            if (thresh[x][y] > 0):
                total_intensity += thresh[x][y]
                pixel_count += 1

    average_intensity = float(total_intensity) / pixel_count
    PEN_PRESSURE = average_intensity
    # print total_intensity
    # print pixel_count
    # print ("Average pen pressure: "+str(average_intensity))

    return


''' main '''


def start(file_name):

    global BASELINE_ANGLE
    global TOP_MARGIN
    global LETTER_SIZE
    global LINE_SPACING
    global WORD_SPACING
    global PEN_PRESSURE
    global SLANT_ANGLE

    # read image from disk
    image = cv2.imread('images/'+file_name)
    # cv2.imshow(file_name,image)

    # Extract pen pressure. It's such a cool function name!
    barometer(image)

    # apply contour operation to straighten the contours which may be a single line or composed of multiple lines
    # the returned image is straightened version of the original image without filtration and binarization
    straightened = straighten(image)
    # cv2.imshow('straightened',straightened)

    # extract lines of handwritten text from the image using the horizontal projection
    # it returns a 2D list of the vertical starting and ending index/pixel row location of each line in the handwriting
    lineIndices = extractLines(straightened)
    # print lineIndices
    # print

    # extract words from each line using vertical projection
    # it returns a 4D list of the vertical starting and ending indices and horizontal starting and ending indices (in that order) of each word in the handwriting
    wordCoordinates = extractWords(straightened, lineIndices)

    # print wordCoordinates
    # print len(wordCoordinates)
    # for i, item in enumerate(wordCoordinates):
    # cv2.imshow('item '+str(i), straightened[item[0]:item[1], item[2]:item[3]])

    # extract average slant angle of all the words containing a long vertical stroke
    extractSlant(straightened, wordCoordinates)

    # cv2.waitKey(0)
    # cv2.destroyAllWindows()

    BASELINE_ANGLE = round(BASELINE_ANGLE, 2)
    TOP_MARGIN = round(TOP_MARGIN, 2)
    LETTER_SIZE = round(LETTER_SIZE, 2)
    LINE_SPACING = round(LINE_SPACING, 2)
    WORD_SPACING = round(WORD_SPACING, 2)
    PEN_PRESSURE = round(PEN_PRESSURE, 2)
    SLANT_ANGLE = round(SLANT_ANGLE, 2)

    return [BASELINE_ANGLE, TOP_MARGIN, LETTER_SIZE, LINE_SPACING, WORD_SPACING, PEN_PRESSURE, SLANT_ANGLE]