Digital-Image-Processing

Various operations for enhancing the images

HW1: Basic Operations

1. Reading and Showing images, understanding data types

• Reading the images in grayscale: cv2.IMREAD_GRAYSCALE
• Changing data type: .astype
• Accessing the pixels through indexing

2. Averaging video sequences

• Reading a video: cv2.VideoCapture()
• Calculating the average and variance of video sequences
• Masking an image

3. Affine transformations

• Defining a function to return an n-bit image
• Using thresholding functions in cv2: cv.THRESH_BINARY, cv.THRESH_BINARY_INV, cv.THRESH_TRUNC, cv.THRESH_TOZERO, cv.THRESH_TOZERO_INV
• Thresholding for segmentation

4. Changing color space and thresholding

• Cropping an image by indexing
• Scale transform: cv.resize()
• Understanding interpolation flags in cv.resize(): cv.INTER_NEAREST, cv.INTER_LINEAR, cv.INTER_AREA, cv.INTER_CUBIC
• Using cv.warpAffine() for affine transformations: shear, translation, rotation

HW2: Intensity-based Operations

1. Histogram Equalization

• Plotting histogram
• Applying power law transform
• Comparing the effect of exponential and logarithmic functions on contrast

2. Transform function for increasing contrast

HW3: Spatial Filtering

1. Edge-detecting filters

• Padding the input image
• Defining mean, median, Sobel, and laplacian kernels

2. Denoising filters

• Local-based filtering
• Intensity-based filtering
• Combining both methods

3. Bit-plane slicing & motion detection

• Bit-wise operations (low significant bits contain details while high significant bits contain the generality of the image)
• Using bit-wise XOR operation to detect the difference between some frames which leads to motion detection

HW4: Frequency Domain Filtering

• Preserving the intensity boundaries during filtering
• Stretching and clipping
• Spatial information in phase image of Fourier transform

HW5: Restoration and Morphology

1. Removing spatial-pattern noise

• Identifying noise in the Fourier transformation of the image
• Applying a notch filter
• Getting inverse Fourier transform

2. Morphology operations for denoising

• Binarizing the image with a specific threshold
• Opening operation (erosion -> dilation -> erosion)
• Closing operation (dilation -> erosion -> dilation)

3. Hole-filling for finding boundaries

• Binarizing the image
• Defining a seed and an ellipse kernel for closing operation
• Defining a function to fill the remaining holes
• Finding the boundaries by subtracting the recent image and its eroded version

HW6: Registration and Segmentation

1. Hough transform for segmentation

• Tuning the thresholds for the canny filter
• Using the mentioned thresholds in the hough transform algorithm

2. Feature-based registration

• Asking the user for entering 3 corresponding points in each image
• Giving the points as input to cv.getAffineTransform() which returns the transformation matrix