Convolution

A standard C++ kernel processor for images

Roadmap

Main Classes

Image

The image class will need to be able represent an arbitraily large image in memory. 
Also, it will need to be able to access pixels by (x,y) coordinates. Ideally, uint_32
with RGBA bit shift will an efficient way to represent colors in memory. This
representation should need only a 1D aggregate (for efficiency). With the following
transformation for an (mxn) image:

Given an index Q:
   Q->(x,y)
   x=Q%n, y=(Q-x)/n

The image class will also need to be able to handle edges. The Image class should be able
to handle coordinates outside an image:

  WRAP:
    Tiles the image by modulation
  MIRROR:
    Flips image tiles by inverse modulation
  EXTEND:
    Extends the edges into the out-of-bounds range
  EMPTY:
    Returns 0 if coordinates falls out of bounds

FileImage

 Given a pointer to an image FileImage will load .ppt, .bmp and .png files into it. Also it
 should be able to save an Image class to disk.

Convolution

  Convolution must be able to hold an mxn convolution in memory with fixed size. Also, it 
  should be able to receive an image pointer and output a kernel process to another image 
  pointer.

Extra Classes

MainWindow

 A window (SDL or Qt) previewer to view images in a viewport.

MainGUI

 An event based gui (librocket/CEGUI/Qt) to load images and interact with a convolution matrix graphically.

Resources

https://en.wikipedia.org/wiki/Kernel_(image_processing)