[openCV, Python] Basic openCV2 usage in python

<Warning>
    This article was written based on my experience.
</Warning>

1. How Read Image in disc?

cv2.imread

You can read by using cv2.imread function.

<Caution>

    In python version 2.7, you need writing path '\\', not '\'

</Caution>

<Code>

import cv2

Image = cv2.imread("Path", cv2.CV_LOAD_IMAGE_GRAYSCALE)

</Code>

2. How Show Image in memory?

cv2.imshow

cv2.waitKey

cv2.destroyWindow

Need cv2.imshow function

<Caution>

    If you not use cv2.waitKey && ( cv2.destroyWindow || cv2.destroyAllWindows ).

    It make python IDE to restart

</Caution>

<Code>

import cv2

def ShowImage(Image, WinName="TempWinName"):

    cv2.imshow(WinName, Image)

    cv2.waitKey()

    cv2.destroyWindow(WinName)

ShowImage(Image)

</Code>

3. How Get Size of Image?

Because openCV2 use numpy's array.

You can get size by using Image.shape

shape == (Row, Col, nChannel) or (Height, Width, nChannel)

<Caution>

    If Image is 3 channel (like RGB or BGR), shape's length is 3

    If Image is grayscale, shape's length is 2

    So if you access nChannel in grayscale, It raiseIndexError

</Caution>

<Code>

</Code>

4. How Access each pixel in Image?

Just you "[]" :)

<Caution>

    Image[☆Height, ☆Width, nChannel]

    If you access all the pixel by using for(;;),

    It will waste long time

    Must use openCV function as possible as you can.

</Caution>

<Code>

</Code>

[openCV, Python] Test Median Blur by using Salt&Pepper Noise

Median Filter : http://en.wikipedia.org/wiki/Median_filter (Wiki)

You can remove drastic noise efficiently.


import cv2
import numpy as np
import random

def MakeSaltAndPepperNoise (Image, SaltNum, PepperNum):
CopyImage = Image.copy()
nChannel = 0

# Get Image size
Width = CopyImage.shape[0]
Height = CopyImage.shape[1]

# If image is grayscale, it not have Image.shape[2]
# so it raise IndexError exception
try:
nChannel = CopyImage.shape[2]
except IndexError:
nChannel = 1

# Make Salt Noise
for Salt in range(0, SaltNum):
# Generate Random Position
RWidth = random.randrange(0, Width)
RHeight = random.randrange(0, Height)
# Make Noise
if nChannel > 1:
for c in range(0, nChannel):
CopyImage[RWidth, RHeight, c] = 255
else:
CopyImage[RWidth, RHeight] = 255

# Make Pepper Noise
for Pepper in range(0, PepperNum):
# Generate Random Position
RWidth = random.randrange(0, Width)
RHeight = random.randrange(0, Height)
# Make Noise
if nChannel > 1:
for c in range(0, nChannel):
CopyImage[RWidth, RHeight, c] = 0
else:
CopyImage[RWidth, RHeight] = 0

return CopyImage

In IDE

NoiseImage

Proccesed Image

[Python] Convert Integer to String

def itoa(Number, Radix=10):
    # Optional code
    if Radix > 16:
        raise ValueError("Radix bigger then 16")

    BASE = "0123456789ABCDEF"
    AlphaList = []

    while Number > 0:
        AlphaList.append(BASE[Number%Radix])
        Number /= Radix

    AlphaList.reverse()

    return "".join(AlphaList)

# in IDE

[openCV, Python] Simple trackbar code for threshold

import cv2
import numpy as np

# Callback Function for Trackbar (but do not any work)
def nothing(*arg):
    pass

# Code here
def SimpleTrackbar(Image, WindowName):
# Generate trackbar Window Name
TrackbarName = WindowName + "Trackbar"

# Make Window and Trackbar
cv2.namedWindow(WindowName)
cv2.createTrackbar(TrackbarName, WindowName, 0, 255, nothing)

# Allocate destination image
Threshold = np.zeros(Image.shape, np.uint8)

# Loop for get trackbar pos and process it
while True:
# Get position in trackbar
TrackbarPos = cv2.getTrackbarPos(TrackbarName, WindowName)
# Apply threshold
cv2.threshold(Image, TrackbarPos, 255, cv2.THRESH_BINARY, Threshold)
# Show in window
cv2.imshow(WindowName, Threshold)

# If you press "ESC", it will return value
ch = cv2.waitKey(5)
if ch == 27:
   break

 cv2.destroyAllWindows()
return Threshold

# in IDE

# Result