Computer Vision

Computer Vision#

  1. OpenCV

back to top

OpenCV#

Image#

  • supported image formats: bmp, pbm, pgm, ppm, sr, ras, jpeg, jpg, jpe, jp2, tiff, tif, png, OpenEXR

  • Image Representation

    • represented as multi-dimensional array, 2D or 3D of numpy.array

    • 1st index: y coordinate or row

    • 2nd index: x coordinate or column

    • 3rd index: color channel

  • Image Manipulation

    • NumPy array operations are useful for image manipulations in OpenCV

    • use OpenCV’s or NumPy’s array slicing to manipulate a large region of an image

    • define ROI (regions of interests), and perform operations on it

    img = cv2.imread("Pic.png")
img[100, 120, 0] = 123  # set single pixel's blue channel
img[:, :, 1] = 0  # set all G values to 0

# copy a portion over to another position
my_roi = img[0:100, 0:100]
img[300:400, 300:400] = my_roi
  • imread(src, mode)

    • load image using the file path, specifying mode is optional

    • return an image in BGR colour format by default

    • example modes: IMREAD_COLOR, IMREAD_UNCHANGED, IMREAD_GRAYSCALE

    • data read is stored in Mat object

  • imwrite(path, src)

    • write to a file

    • requires src to be in BGR or grayscale format that the output format can support

  • samples.findFile(): images from OpenCV samples

  • imshow(title, Mat_obj)

    • show image

    • waitKey(time): wait for user input, and return the value of the key pressed

  • cvtColor(src, code, dst)

    • convert image to another colour space

Mat Class#

  • Matrix Header

    • contain information such as the size of the matrix, storing method, and matrix address

    • constant size

    • can have headers which refer to only a subsection of the full data

  • Pointer

    • points to the matrix with the pixel values

    • matrix size varies from image to image

  • each Mat object has own header, but a matrix may be shared by having same pointer values

  • assignment and copy operators only copy the headers and the pointer

  • the last object that used the Mat object cleans it up

  • Mat_OBJ.clone()

    • clone the matrix

  • Mat_OBJ.copyTo(Mat_OBJ)

    • copy the matrix to another object

back to top