Repo for assignments done in ELL715: Digital Image Processing @ IIT Delhi
In each assignment folder, there's the problem statement, the code and the final report submitted.
The main tasks performed in the assignments are:
- Image Transformations:
- Polar transformation
- Shear transformation
- Geometric rotation, scaling and translation
- Image Histogram and Equalization:
- Computed histogram of an RGB image
- Improved contrast of a grayscale image using:
- Contrast Stretching
- Gamma Correction
- Histogram Equalization
- Unsharp masking
- Image Filtering and Scaling:
- Analyzed the effect of an IIR filter and unsharp masking (shaprening filter)
- Played with fourier transform and inverse fourier transform
- Performed Discrete Cosine Transform on images
- Image Compression:
- Implemented bilinear, bicubic, DCT and FFT compression
- Implemented the Seam carving by removing areas with low energy
- Impelemnted compression using 2-D Wavelet decomposition
- Wavelet Analysis:
- Extracted retinal blood vessels from fundus images using:
- Wavelets
- Sobel
- LoG
- Extracted retinal blood vessels from fundus images using:
- Image Morphing:
- Separated occluded objects using morphology and thresholding
- Counted circular objects in an image
- Detected black note heads
- Vessel Segmentation using Morphology:
- Removed the bright, circular shaped Optic Disc
- Extracted the vessel map
- Estimated the diameter of thinnest & thickest vessel
- Removed small isolated patches and smoothened the vessel map
- Image Segmentation:
- Implemented segmentation using Intensity based thresholding
- Implemented segmentation using Region growing
- Implemented segmentation using Image Matting
- Optic Disc Segmentation Using Active Contour:
- Imeplemented algorithm to find the center of the optic disc
- Developed an active contour algorithm to fit the initial contour at the edges of the disc
- Removed small isolated patches and smoothened the vessel map
- Retinal Vessel Segmentation Using Neural Networks:
- Extracted features for vessel map segmentation
- Followed leave-one-out strategy to train the neural network
- Evaluated the performance of algorithm in terms of:
- Accuracy
- Sensitivity
- Specificity
- Area Under Receiver Operating Characteristics Curve (AUC under ROC)
Most of the work here was done in collaboration with @naman1122000