SBS to Anaglyph

Explination

This program converts a single image with side by side left and right views Into an anaglyph (red blue) image by finding the best transformation to overlay them
using computer vision techniques from
simultanious localization and mapping,
structure from motion (3d view from video)
such as
edge detection,
maximum local feature detection,
correspondance matching
to automatically
align the images as best as possible

to generate an anaglyph from left and right images
what is acutally needed is the lens transformation for both images
is many image pairs with correlated features at different depths across the frames
then given the sbs lens doesn't move, the static transform can be found

this program doesn't do that yet
and is more of a test program for the feature detection and matching
functions that are planned to be the basis of a program library to transform
video and images into 3d objects and labeled components for video games and robots
to be able to understand and interact with the world
once it is working reliably with a stereo pair of a known translation
then it should be able to find the transformation between two video frames with static/fixed postion objects
and be used to generate a camera path in 3d and build a 3d pointcloud from feature points

Settings

Feature Correspondance Match Threshold
Max Number of Correspondence Sets How many subsets of the corresponding features to generate and choose from
Feature Detection Pixel Variance Threshold The amount a pixel in the edge detected downsampled image must differ from 127 (neutral gray) for it to be considered a feature
Max Vert Diff The percentage of image height to search vertically for corresponding/matching features
Expected Horiz Diff The percentage of image width to search horizontally for corresponding/matching features
Max Horiz Diff The percentage of image width to search horizontally for corresponding/matching features
Num Times to Downsample before feature detection 1 or higher reccomended for performance and reducing pixel noise sensitivity, (this is the number times the image resolution is divided by 2), too low and memory and compute time may be too high
Input Image
Run	Set Defaults