trangnm58/opencv

Repository for learning OpenCV (Python)

##Requirements:

• Python 2.7 with numpy and matplotlib
• OpenCV 3.1

##Usage: ###draw_canvas.py

python draw_canvas.py

###merge_images.py

python merge_images.py

(put OpenCV logo 'thing.png' on top of my picture 'some.png')

##OpenCV summary

1. GUI features (draw_canvas.py)

• Read, show, write image, video
• Draw line, circle, rectangle, ellipse, polygon
• Add text to image, video
• Mouse events
• Trackbar

2. Core operations (merge_images.py)

• ROI (Region of Image)
• Image properties (shape, size)
• Split and Merge image channels (BGR image => B, G, R)
• Make borders for image
• Add images
• Image blending: g(x) = (1-a)f0(x) + af1(x)
• Bitwise operations: and, or, not, xor

3. Image processing (video/video.py)

• Change colorspaces (BGR => grayscale or BGR => HSV)
• Image Thresholding
• Geometric Transformations (move, scale, rotate)
  • Affine Transformation (2D): 2x3 matrix
  • Perspective Transformation (3D): 3x3 matrix
• Smooth image (blur, denoise):
  • 2D convolution, filter, kernel
  • LPF removes noise, HPF detects edges
  • Averaging, Gaussian, Median, Bilateral (preserve edges)
• Morphological Transformations: + Erosion removes noise but shrinks things, Dilation makes them big again
  • Opening = erosion + dilation
  • Closing = dilation + erosion
  • Morphological Gradient = dilation - erosion
  • Top Hat = origin - opening
  • Black Hat = closing - origin
• Image Gradients
• Canny Edge Dectection
• Hough Line, Circle Transform
• Image Pyramids: different resolutions of a image stack together
• Contours: boundaries of a shape with same intensity
• Histograms: intensity distribution of an image
  • Histogram equalization: improve the contrast
• Fourier Transform
• Template Matching

4. Feature Detection
5. Video Analysis (video/object_tracking.py)

• Object tracking
• Optical Flow
• Background Subtraction

6. Camera Calibration and 3D Reconstruction
7. Machine Learning

• kNN
• SVM
• k-Means

8. Computational Photography

• Image Denoising: average similar patches in different places
• Image Inpainting