Part I: Image Cartoonifying

This part explores using various image processing filters to transform real-world images into a cartoon-like style.

Key Steps

• Black and White Sketch:
  • Noise Reduction: Median filter cleans the image.
  • Edge Detection: Laplacian filter identifies edges, converted to black/white using thresholding.
• Color Painting:
  • Bilateral Filtering: Smooths flat regions while preserving edges.
  • Optimization: Apply smaller, truncated filters for efficiency and similar effect.
• Cartoon Creation:
  • Overlay the black and white sketch onto the color painting, preserving edges and creating the cartoon effect.

Results

Original Image

Noise Reduction

Edge Detection

Thresholding

Smoothing

Cartoonified

Part II: Road Lane Detection with Hough Transform

This part explores using the Hough Transform to detect road lanes within an image.

Key Steps:

• Image Preprocessing:

  • Smoothing: A 2D median filter removes noise.
  • Edge Detection: Canny edge detection with high thresholds identifies strong edges.

• Region of Interest (ROI):

  • A defined polygon masks irrelevant edges, focusing on the potential road area.

• Hough Transform:

  • Accumulates votes for potential line parameters (ρ, θ) based on edge points.
  • Identifies peaks in the accumulator array as likely lane lines.

• Refinement:

  • Non-maximum suppression removes weaker lines close to stronger ones.

Results:

Original Image

Masked Edges

Accumulator array

Road with lanes