YFilterSize = Statistics.RoundToOdd(len(img) / 20)
XFilterSize = Statistics.RoundToOdd(len(img[0]) / 20)
SmoothSize = Statistics.RoundToEven(len(img) / 20)
print "smooth size", SmoothSize
# use the recomended standard deviation
blur = cv2.GaussianBlur(img, (XFilterSize, YFilterSize), 0)
fig, ((ax1, ax2), (ax3, ax4), (ax5, ax6)) = plt.subplots(3,
2,
sharex=False,
sharey=False)
BlurredImageArray = Array.VerticalArrayFromImage(Image.fromarray(blur),
BandPercentage=.5)
MeanArray = Array.MeanArray(BlurredImageArray, SmoothSize)
MedianArray = Array.MedianArray(BlurredImageArray, SmoothSize)
MeanMedianArray = Array.MedianArray(MeanArray, SmoothSize)
ax1.plot(MeanMedianArray)
ax1.set_title("Blurred then Mean then Median")
ax2.plot(BlurredImageArray)
ax2.set_title("Blurred")
ax3.plot(MeanArray)
ax3.set_title("Blurred and Mean")
ax4.plot(Array.VerticalArrayFromImage(Image.fromarray(img), BandPercentage=.5))
ax4.set_title("unprocessed")
ax5.plot(MedianArray)
ax5.set_title("Blurred then Median")
plt.show()
# get the line height
LineHeight = 0
with open('SharedData/LineHeight.txt', 'r') as content_file:
LineHeight = int(float(content_file.read()))
# use the line height, and make the smoothing size 1/2 of that
SmoothSize = Statistics.RoundToEven(LineHeight / 2.0)
print "smooth size", SmoothSize
# use the recomended standard deviation
blur = cv2.GaussianBlur(img, (XFilterSize, XFilterSize), 0)
BlurredImageArray = Array.HorizontalArrayFromImage(Image.fromarray(blur))
RegularImageArray = Array.HorizontalArrayFromImage(Image.fromarray(img))
MeanArray = Array.MeanArray(RegularImageArray, SmoothSize)
MeanArrayBlurred = Array.MeanArray(BlurredImageArray, SmoothSize)
fig, ((ax1, ax2)) = plt.subplots(2, 1, sharex=False, sharey=False)
# LocalMaxes = sorted(Array.FindAllLocalMaxes(BlurredImageArray) + Array.FindAllLocalMins(BlurredImageArray))
LocalMaxes = Array.FindAllLocalMaxes(BlurredImageArray)
# sets how different maximums from the 'same' point can be
# this may be incorrect and will F**K us
NewMinimumThreshold = 5
# make sure that clusters of data are treated as one individual point
FilteredLocalMaxes = []
# add the first point because it automatically counts
FilteredLocalMaxes.append(LocalMaxes[0])
def CalculateCropPoints(LineHeight, ShowGraphs=False):
global FileName
# do the gaussian smoothing that enables us to find the crop points
# blur the image
img = cv2.imread("Processed/" + FileName)
# compute the raw array
RawArray = Array.VerticalArrayFromImage(Image.fromarray(img))
# calculate filter and smooth sizes based on how big the image is
YFilterSize = Statistics.RoundToOdd(len(img) / 20)
XFilterSize = Statistics.RoundToOdd(len(img[0]) / 20)
SmoothSize = Statistics.RoundToEven(len(img) / 20)
# use the line height, and make the smoothing size 1/2 of that
SmoothSize = Statistics.RoundToEven(LineHeight / 2.0)
# use the recomended standard deviation
blur = cv2.GaussianBlur(img, (XFilterSize, YFilterSize), 0)
BlurredImageArray = Array.VerticalArrayFromImage(Image.fromarray(blur),
BandPercentage=.5)
MeanArray = Array.MeanArray(BlurredImageArray, SmoothSize)
MedianArray = Array.MedianArray(BlurredImageArray, SmoothSize)
# get the crop points by finding local maxes
CropPoints = Array.FindAllLocalMaxes(MeanArray)
# check to see if there should be any extra crop points added
SlopeArray = Array.FindSlope(MeanArray)
# if the initial slope is less than 0
if SlopeArray[0] < 0:
# insert 0 at position 0
CropPoints.insert(0, 0)
# if the final slope is more than 0 or 0
if SlopeArray[len(SlopeArray) - 1] >= 0:
CropPoints.append(len(MeanArray) - 1)
if ShowGraphs:
# graph / visualize some stuff
TempArray = []
for i in range(0, len(MeanArray)):
if i in CropPoints:
TempArray.append(255)
else:
TempArray.append(150)
fig, ((ax1, ax2), (ax3, ax4)) = plt.subplots(2,
2,
sharex=False,
sharey=False)
ax1.set_title("Blurred then Mean")
ax1.plot(MeanArray)
ax1.plot(RawArray)
ax2.plot(BlurredImageArray)
ax2.plot(RawArray)
ax2.set_title("Blurred")
ax3.imshow(img)
ax3.set_title("Image")
ax4.imshow(blur)
ax4.set_title("Blur Image")
plt.show()
return CropPoints