supper/src/ocr.rs

use image::{Rgb, RgbImage};
use img_hash::ImageHash;

use crate::image_processing;

#[derive(Debug)]
struct BoundingBox {
    x: u32,
    y: u32,
    width: u32,
    height: u32,
}

fn column_has_any_dark(image: &RgbImage, x: u32) -> bool {
    for y in 0..image.height() {
        let [r, g, b] = image.get_pixel(x, y).0;
        if r < 100 && g < 100 && b < 100 {
            return true;
        }
    }
    false
}

fn row_has_any_dark(image: &RgbImage, y: u32, start_x: u32, width: u32) -> bool {
    for x in start_x..(start_x + width) {
        let [r, g, b] = image.get_pixel(x, y).0;
        if r < 100 && g < 100 && b < 100 {
            return true;
        }
    }
    false
}

fn take_while<F: Fn(u32) -> bool>(x: &mut u32, max: u32, f: F) {
    while *x < max && f(*x) {
        *x += 1;
    }
}

fn get_character_bounding_boxes(image: &RgbImage) -> Vec<BoundingBox> {
    let mut x = 0;
    let mut boxes = Vec::new();
    while x < image.width() {
        take_while(&mut x, image.width(), |x| !column_has_any_dark(image, x));

        let start_x = x;
        take_while(&mut x, image.width(), |x| column_has_any_dark(image, x));
        let width = x - start_x;

        if width >= 1 {
            let mut y = 0;
            take_while(&mut y, image.height(), |y| {
                !row_has_any_dark(image, y, start_x, width)
            });

            let start_y = y;

            let mut inverse_y = 0;
            take_while(&mut inverse_y, image.height(), |y| {
                !row_has_any_dark(image, image.height() - 1 - y, start_x, width)
            });
            let end_y = image.height() - inverse_y;
            let height = end_y - start_y;
            if height >= 1 {
                boxes.push(BoundingBox {
                    x: start_x,
                    y: start_y,
                    width,
                    height,
                });
            }
        }
    }
    boxes
}

fn trim_to_bounding_box(image: &RgbImage, bounding_box: &BoundingBox) -> RgbImage {
    const PADDING: u32 = 2;
    let mut buffer = RgbImage::from_pixel(
        bounding_box.width + 2 * PADDING,
        bounding_box.height + 2 * PADDING,
        Rgb([0xFF, 0xFF, 0xFF]),
    );
    for y in 0..bounding_box.height {
        for x in 0..bounding_box.width {
            buffer.put_pixel(
                x + PADDING,
                y + PADDING,
                *image.get_pixel(bounding_box.x + x, bounding_box.y + y),
            );
        }
    }
    buffer
}

pub fn bounding_box_images(image: &RgbImage) -> Vec<RgbImage> {
    let mut trimmed = Vec::new();

    let boxes = get_character_bounding_boxes(image);
    for bounding_box in boxes {
        trimmed.push(trim_to_bounding_box(image, &bounding_box));
    }
    trimmed
}

pub fn compute_box_hashes(image: &RgbImage) -> Vec<ImageHash> {
    let mut hashes = Vec::new();

    let boxes = get_character_bounding_boxes(image);
    for bounding_box in boxes {
        let trimmed = trim_to_bounding_box(image, &bounding_box);
        hashes.push(image_processing::hash_image(&trimmed))
    }
    hashes
}

#[test]
fn test_bounding_boxes() {
    let image_bytes = include_bytes!("test_data/test-image-2.png");
    let image = image::load_from_memory(image_bytes).unwrap().to_rgb8();
    let boxes = get_character_bounding_boxes(&image);
    assert_eq!(boxes.len(), 10);
    assert_ne!(boxes[0].x, 0);
    assert_ne!(boxes[0].y, 0);
    assert_ne!(boxes[0].height, 0);
    assert_ne!(boxes[0].width, 0);
}

#[test]
fn test_box_hashes() {
    let image_bytes = include_bytes!("test_data/test-image-2.png");
    let image = image::load_from_memory(image_bytes).unwrap().to_rgb8();
    let hashes = compute_box_hashes(&image);
    assert_eq!(hashes.len(), 10);
}