USC EE569 - Introduction to Image Processing 20 Spring

Type: Course Work

Languange:

• C++11 (Major Language)
• Python (Subsidiary script for validating numerical outputs & Visualization);

For image results, please refer to the PDF reports.

Related Algorithms & Topics:

• Assignment 1:
  • Demosaicing:
    • Bilinear interpolation
    • Malvar-He-Cutler interpolation
  • Brightness Enhancement (Histogram Manipulation):
    • Transfer function based
    • Cumulative Probability based
  • Denoising:
    • Uniform Kernel, Gaussain Kernel
    • Bilateral Kernel
    • Non-Local-Means (self-implemented), BM3D (OpenCV)
• Assignment 2:
  • Edge Detection:
    • Sobel edge detectors (self-implemented)
    • Canny edge detectors (OpenCV)
    • Structured edge detectors (OpenCV)
    • F1-score calculation for edge detectors (self-implemented)
  • Digital Half-Toning:
    • Dithering
      • Naive Thresholding (Fixed T & Uniform Random T)
      • Dithering index matrix (Shifting Mask)
    • Error Diffusion (Sepentime Traversal):
      • Floyd-Steinberg's, JJN's, Stucki's error diffusion matrix / kernel.
      • Gray scale images, colored images by seperate diffusion & MBVQ-based diffusion.
• Assignment 3:
  • Geometric Transformation
  • Affine & Projective Transfomration
    • Image Stitching Using SURF+FLANN for control point detection
  • Binary Image Morphological Transformation
    • Thinning, Shrinking, Skeletonizing
    • Star number counting, star size counting
    • PCB analysis, detecting wires & holes
    • Defection detection & completion
  • Additional works:
    • Matrix calculation Toolbox
      • Matrix allocation, Mat-Mat/Vec-Mat/Mat multiplication, transpose
• Assignment 4:
  • (Image Based) Texture Classfication
    • Lowe's Filters used for feature extraction
    • Implemented ML Algorithm: K-Means (Naive Start & K-Means++), PCA
    • Called ML Algorithm: SVM / Kernel Machine, Randorm Forest
  • Texture Segmentation
    • Lowe's Filter + K-Means
  • SIFT feature extraction & Feature Matching
  • Additional Works:
    • Utilization of data structures in std (std::Vector)
    • API encapsulation & OOP programming
    • Refinement of Matrix_ToolBox / IO functions / Image Operations.
• Assignment 5:
  • Convolutional Neural Network Training
  • Model: LeNet5, ResNetv1 (for CIFAR10 )
  • Dataset: CIFAR10
  • Additional Works:
    • Configuration & Progress Recorder in JSON format
    • Replicating Famous CNNs: SqueezeNet, MobileNetv1 & Network In Network
  • Follow up please refer to my other Repo.
• Assignment 6:
  • Subspace Sucessive Learning (SSL) for image classification
  • Dataset: CIFAR10
  • Additional Works:
    • Multi-threading & Multi-processing in Python
    • Use of Google Cloud Platform

Dependencies:

OpenCV C++ Library / Eigen3 Library