docs: write README

README.md

@@ -3,3 +3,50 @@
 **Convert diagram pictures into SVGs**
 
 ![image info](./banner.png)
+
+## What is this?
+This is a tool that allows converting a JPEG photo, e.g. a diagram photo,
+into an SVG picture, as seen in the example abvove.
+
+## How do I use it?
+As seen in [examples/main.rs](./examples/main.rs):
+```rs
+use u_u::jpeg_to_svg;
+
+const JPEG_BYTES: &[u8] = include_bytes!("./u_u.jpg");
+
+fn main() {
+    let svg_bytes = jpeg_to_svg(JPEG_BYTES).unwrap();
+    std::fs::write("./output_example.svg", svg_bytes).unwrap();
+}
+```
+
+### How does it work?
+The algorithm interprets each RGB pixel as a 3D vector.
+
+The first step is to calculate the background colour, which is calculated as
+the average colour of all the pixels in the image -- this means that the
+background of the photo (such as the white of a whiteboard, or black of a
+blackboard) must be more than 50% of the photo, or this fails (it could be
+an optional argument to pass the background colour if it is known to
+prevent failure on those cases).
+
+Then, there are two passes on the image to filter out the background:
+
+1. The first pass tries to identify what is part of the foreground, and removes
+the background parts according to a threshold value.
+This results in a new image.
+2. The second pass tries to identify what is part of the background, and removes
+the foreground parts according to a threshold value.
+This results in a new image.
+3. The images from 1. and 2. are merged.
+All second pass pixels (background) are removed from the foreground.
+Non-second-pass pixels that are similar to the foreground colour, and
+different to the background colour, are added to the foreground if not already
+in the foreground -- this reduces errors from first pass.
+This cleans the foreground of any background that wasn't removed in the first pass.
+4. Pixels with a low neighbour count (i.e, surrounded by many transparent
+pixels) are removed. This further removes noise from the picture.
+5. The resulting RBG bytes are converted to SVG using [vtracer][0].
+
+[0]: https://github.com/visioncortex/vtracer