Deimos

Deimos is an experimental computer vision library in C

I am trying to distill my knowledge into a computer vision guide, which serves as an introductory guide to this field and goes into details about the algorithms implemented below. I welcome any suggestions/critiques, thanks!

Features

• Geometric Transformations
  • Rotation
  • Scaling
  • Shearing/Warping
• Smoothing
  • Gaussian filter
  • Median filter
  • Bilateral filter
  • Box filter
• Edge detection
  • Unsharp mask
  • Laplacian Filter (Difference of Gaussian)
  • Sobel operator
• Color
  • Greyscale
  • Contrast
• Thresholding
  • Global
  • Otsu's Method
• Feature extraction
  • Harris corner detection

Build (Make)

git clone git@github.com:aadv1k/deimos
cd deimos/
make

Build (without make)

If you don't have make on windows, you can instead use the build.bat, it gets the job done.

git clone git@github.com:aadv1k/deimos
cd deimos/
.\build.bat

Examples

Note The filters are un-optimized, they work best on smaller images

.\cv gray ..\data\img1.jpg .\output.png
.\cv sharpen --sigma 3 ..\data\img1.jpg .\output.png
.\cv blur --kernel 9 --sigma 5 ..\data\img1.jpg .\output.png

Usage

Usage:
  C:\Users\Aadv1k\Desktop\deimos\bin\deimos.exe <command> <args> input output

Examples:
  deimos blur --sigma 3.4 --kernel 5 input.png output.png
  deimos gray input.png output.png
  deimos median --kernel 3 input.png output.png
  deimos sharpen --sigma 0.6 --kernel 3 input.png output.png

Commands:
  Smoothing:
    blur, gaussian    Apply Gaussian blur to the image.
    median            Apply median filter to the image.
    bilateral         Apply bilateral filter to the image.
    box, mean         Apply box filter onto the image.

  Edge Detection:
    sobel             Apply Sobel filter to the image.
    laplacian         Apply Laplacian filter onto the image.
    sharpen           Sharpen image via an unsharp mask.

  Thresholding:
    global-threshold  Apply the global threshold filter over the image. 'sigma' would be the threshold.
    otsu-threshold    Apply Otsu's threshold filter over the image.

  Feature Extraction:
    harris-corners    Detect corners within the image via Harris corner detection.

  Transformations:
    rotate            Rotate the image by `sigma` deg.
    scale             Scale the image by sigma factor; if sigma is negative then downscale else upscale
    shear             Shear or skew the image at both axis by sigma factor.
    flipX             Horizontally mirror the image
    flipY             Vertically mirror the image

Color enhancements:
    gray              Convert image to grayscale.
    contrast          Increase/decrease contrast based on `sigma`

Other Commands:
    help              Print this help message.

Options:
  --sigma             Specify the primary modifier for the convolutions.
  --kernel            Define the kernel size for convolutions (if applicable).
  --no-threshold      Disable specifying a threshold for the sobel operator, default to setting gradient magnitude 

Gallery

Note this showcases only some of the many avaliable filters

Gaussian blur

.\bin\deimos blur --kernel 9 --sigma 3 .\data\img1.jpg .\img1-gaussian-3-9.png

Original Image	Gaussian Blur

Median filter

.\bin\deimos median --kernel 9 .\data\img1.jpg ..\output.jpg

Original Image	Median Filter

Unsharp mask

.\bin\deimos sharpen --kernel 9 --sigma 1.2 .\data\img1.jpg ..\output.jpg

Original Image	Unsharp Mask

Laplacian filter

.\bin\deimos laplacian --kernel 3 --sigma 1.5 .\data\img1.jpg .\output.jpg

Original Image	Laplacian Filter

Otsu's Threshold

.\bin\deimos sobel otsu-threshold .\data\img1.jpg .\output.jpg

Original Image	Otsu's threshold

Sobel operator

.\bin\deimos sobel --no-threshold .\data\img1.jpg .\output.jpg

Original Image	Sobel Operator