Update README.md

added #[derive(Clone, Debug)]

.github/workflows/rust.yml

@@ -1,10 +1,23 @@
 name: Rust
 
 on:
-  push:
-    branches: [ master ]
   pull_request:
-    branches: [ master ]
+    paths-ignore:
+      - '**.md'
+      - '.github/ISSUE_TEMPLATE/**'
+  push:
+    paths-ignore:
+      - '**.md'
+      - '.github/ISSUE_TEMPLATE/**'
+    branches:
+      - master
+      - 0.*.x
+      - pr/**/ci
+      - ci-*
+
+concurrency:
+  group: ${{ github.workflow }}-${{ github.head_ref || github.ref || github.run_id }}
+  cancel-in-progress: true
 
 env:
   CARGO_TERM_COLOR: always

LICENSE

@@ -1,4 +1,4 @@
-Copyright (c) 2023 Tsang Hao Fung
+Copyright (c) 2024 TSANG, Hao Fung
 
 Permission is hereby granted, free of charge, to any
 person obtaining a copy of this software and associated

README.md

@@ -10,9 +10,9 @@
   <h3>
     <a href="https://www.visioncortex.org/vtracer-docs">Article</a>
     <span> | </span>
-    <a href="https://www.visioncortex.org/vtracer/">Demo</a>
+    <a href="https://www.visioncortex.org/vtracer/">Web App</a>
     <span> | </span>
-    <a href="https://github.com/visioncortex/vtracer/releases/tag/0.6.0">Download</a>
+    <a href="https://github.com/visioncortex/vtracer/releases">Download</a>
   </h3>
 
   <sub>Built with 🦀 by <a href="https://www.visioncortex.org/">The Vision Cortex Research Group</a></sub>
@@ -28,7 +28,7 @@ Comparing to Adobe Illustrator's [Image Trace](https://helpx.adobe.com/illustrat
 
 VTracer is originally designed for processing high resolution scans of historic blueprints up to gigapixels. At the same time, VTracer can also handle low resolution pixel art, simulating `image-rendering: pixelated` for retro game artworks.
 
-A technical description of the algorithm is on [visioncortex.org/vtracer-docs](https://www.visioncortex.org/vtracer-docs).
+Technical descriptions of the [tracing algorithm](https://www.visioncortex.org/vtracer-docs) and [clustering algorithm](https://www.visioncortex.org/impression-docs).
 
 ## Web App
 
@@ -71,9 +71,9 @@ OPTIONS:
     -s, --splice_threshold <splice_threshold>    Minimum angle displacement (degree) to splice a spline
 ```
 
-### Install
+## Downloads
 
-You can download pre-built binaries from [Releases](https://github.com/visioncortex/vtracer/releases/tag/0.6.0).
+You can download pre-built binaries from [Releases](https://github.com/visioncortex/vtracer/releases).
 
 You can also install the program from source from [crates.io/vtracer](https://crates.io/crates/vtracer):
 
@@ -81,13 +81,15 @@ You can also install the program from source from [crates.io/vtracer](https://cr
 cargo install vtracer
 ```
 
+> You are strongly advised to not download from any other third-party sources 
+
 ### Usage
 
 ```sh
 ./vtracer --input input.jpg --output output.svg
 ```
 
-## Rust Library
+### Rust Library
 
 You can install [`vtracer`](https://crates.io/crates/vtracer) as a Rust library.
 
@@ -95,7 +97,7 @@ You can install [`vtracer`](https://crates.io/crates/vtracer) as a Rust library.
 cargo add vtracer
 ```
 
-## Python Library
+### Python Library
 
 Since `0.6`, [`vtracer`](https://pypi.org/project/vtracer/) is also packaged as Python native extensions, thanks to the awesome [pyo3](https://github.com/PyO3/pyo3) project.
 
@@ -105,7 +107,7 @@ pip install vtracer
 
 ## In the wild
 
-VTracer is used by the following projects (feel free to add yours!):
+VTracer is used by the following products (open a PR to add yours):
 
 <table>
   <tbody>
@@ -118,12 +120,20 @@ VTracer is used by the following projects (feel free to add yours!):
   </tbody>
 </table>
 
-## Anecdotes
+## Citations
+
+VTracer has since been cited by a few academic papers in computer graphics / vision research. Please kindly let us know if you have cited our work:
+
++ SKILL 2023 [Framework to Vectorize Digital Artworks for Physical Fabrication based on Geometric Stylization Techniques](https://www.researchgate.net/publication/374448489_Framework_to_Vectorize_Digital_Artworks_for_Physical_Fabrication_based_on_Geometric_Stylization_Techniques)
++ arXiv 2023 [Image Vectorization: a Review](https://arxiv.org/abs/2306.06441)
++ arXiv 2023 [StarVector: Generating Scalable Vector Graphics Code from Images](https://arxiv.org/abs/2312.11556)
++ arXiv 2024 [Text-Based Reasoning About Vector Graphics](https://arxiv.org/abs/2404.06479)
++ arXiv 2024 [Delving into LLMs' visual understanding ability using SVG to bridge image and text](https://openreview.net/pdf?id=pwlm6Po61I)
+
+## How did VTracer come about?
 
 > The following content is an excerpt from my [unpublished memoir](https://github.com/visioncortex/memoir).
 
-### How / when / where did VTracer come about?
-
 At my teenage, two open source projects in the vector graphics space inspired me the most: Potrace and Anti-Grain Geometry (AGG).
 
 Many years later, in 2020, I was developing a video processing engine. And it became evident that it requires way more investment to be commercially viable. So before abandoning the project, I wanted to publish *something* as open-source for posterity. At that time, I already developed a prototype vector graphics tracer. It can convert high-resolution scans of hand-drawn blueprints into vectors. But it can only process black and white images, and can only output polygons, not splines.

cmdapp/pyproject.toml

@@ -1,6 +1,6 @@
 [project]
 name = "vtracer"
-version = "0.6.10"
+version = "0.6.11"
 description = "Python bindings for the Rust Vtracer raster-to-vector library"
 authors = [ { name = "Chris Tsang",  email = "chris.2y3@outlook.com" } ]
 readme = "vtracer/README.md"

cmdapp/src/config.rs

@@ -1,23 +1,27 @@
 use std::str::FromStr;
 use visioncortex::PathSimplifyMode;
 
+#[derive(Debug, Clone)]
 pub enum Preset {
     Bw,
     Poster,
     Photo,
 }
 
+#[derive(Debug, Clone)]
 pub enum ColorMode {
     Color,
     Binary,
 }
 
+#[derive(Debug, Clone)]
 pub enum Hierarchical {
     Stacked,
     Cutout,
 }
 
 /// Converter config
+#[derive(Debug, Clone)]
 pub struct Config {
     pub color_mode: ColorMode,
     pub hierarchical: Hierarchical,
@@ -32,6 +36,7 @@ pub struct Config {
     pub path_precision: Option<u32>,
 }
 
+#[derive(Debug, Clone)]
 pub(crate) struct ConverterConfig {
     pub color_mode: ColorMode,
     pub hierarchical: Hierarchical,

cmdapp/src/converter.rs

@@ -73,7 +73,7 @@ fn should_key_image(img: &ColorImage) -> bool {
 
     // Check for transparency at several scanlines
     let threshold = ((img.width * 2) as f32 * KEYING_THRESHOLD) as usize;
-    let mut num_transparent_boundary_pixels = 0;
+    let mut num_transparent_pixels = 0;
     let y_positions = [
         0,
         img.height / 4,
@@ -84,9 +84,9 @@ fn should_key_image(img: &ColorImage) -> bool {
     for y in y_positions {
         for x in 0..img.width {
             if img.get_pixel(x, y).a == 0 {
-                num_transparent_boundary_pixels += 1;
+                num_transparent_pixels += 1;
             }
-            if num_transparent_boundary_pixels >= threshold {
+            if num_transparent_pixels >= threshold {
                 return true;
             }
         }

cmdapp/src/python.rs

@@ -1,5 +1,7 @@
 use crate::*;
-use pyo3::prelude::*;
+use image::{io::Reader, ImageFormat};
+use pyo3::{exceptions::PyException, prelude::*};
+use std::io::{BufReader, Cursor};
 use std::path::PathBuf;
 use visioncortex::PathSimplifyMode;
 
@@ -23,6 +25,148 @@ fn convert_image_to_svg_py(
     let input_path = PathBuf::from(image_path);
     let output_path = PathBuf::from(out_path);
 
+    let config = construct_config(
+        colormode,
+        hierarchical,
+        mode,
+        filter_speckle,
+        color_precision,
+        layer_difference,
+        corner_threshold,
+        length_threshold,
+        max_iterations,
+        splice_threshold,
+        path_precision,
+    );
+
+    convert_image_to_svg(&input_path, &output_path, config).unwrap();
+    Ok(())
+}
+
+#[pyfunction]
+fn convert_raw_image_to_svg(
+    img_bytes: Vec<u8>,
+    img_format: Option<&str>, // Format of the image (e.g. 'jpg', 'png'... A full list of supported formats can be found [here](https://docs.rs/image/latest/image/enum.ImageFormat.html)). If not provided, the image format will be guessed based on its contents.
+    colormode: Option<&str>,  // "color" or "binary"
+    hierarchical: Option<&str>, // "stacked" or "cutout"
+    mode: Option<&str>,       // "polygon", "spline", "none"
+    filter_speckle: Option<usize>, // default: 4
+    color_precision: Option<i32>, // default: 6
+    layer_difference: Option<i32>, // default: 16
+    corner_threshold: Option<i32>, // default: 60
+    length_threshold: Option<f64>, // in [3.5, 10] default: 4.0
+    max_iterations: Option<usize>, // default: 10
+    splice_threshold: Option<i32>, // default: 45
+    path_precision: Option<u32>, // default: 8
+) -> PyResult<String> {
+    let config = construct_config(
+        colormode,
+        hierarchical,
+        mode,
+        filter_speckle,
+        color_precision,
+        layer_difference,
+        corner_threshold,
+        length_threshold,
+        max_iterations,
+        splice_threshold,
+        path_precision,
+    );
+    let mut img_reader = Reader::new(BufReader::new(Cursor::new(img_bytes)));
+    let img_format = img_format.and_then(|ext_name| ImageFormat::from_extension(ext_name));
+    let img = match img_format {
+        Some(img_format) => {
+            img_reader.set_format(img_format);
+            img_reader.decode()
+        }
+        None => img_reader
+            .with_guessed_format()
+            .map_err(|_| PyException::new_err("Unrecognized image format. "))?
+            .decode(),
+    };
+    let img = match img {
+        Ok(img) => img.to_rgba8(),
+        Err(_) => return Err(PyException::new_err("Failed to decode img_bytes. ")),
+    };
+    let (width, height) = (img.width() as usize, img.height() as usize);
+    let img = ColorImage {
+        pixels: img.as_raw().to_vec(),
+        width,
+        height,
+    };
+    let svg =
+        convert(img, config).map_err(|_| PyException::new_err("Failed to convert the image. "))?;
+    Ok(format!("{}", svg))
+}
+
+#[pyfunction]
+fn convert_pixels_to_svg(
+    rgba_pixels: Vec<(u8, u8, u8, u8)>,
+    size: (usize, usize),
+    colormode: Option<&str>,       // "color" or "binary"
+    hierarchical: Option<&str>,    // "stacked" or "cutout"
+    mode: Option<&str>,            // "polygon", "spline", "none"
+    filter_speckle: Option<usize>, // default: 4
+    color_precision: Option<i32>,  // default: 6
+    layer_difference: Option<i32>, // default: 16
+    corner_threshold: Option<i32>, // default: 60
+    length_threshold: Option<f64>, // in [3.5, 10] default: 4.0
+    max_iterations: Option<usize>, // default: 10
+    splice_threshold: Option<i32>, // default: 45
+    path_precision: Option<u32>,   // default: 8
+) -> PyResult<String> {
+    let expected_pixel_count = size.0 * size.1;
+    if rgba_pixels.len() != expected_pixel_count {
+        return Err(PyException::new_err(format!(
+            "Length of rgba_pixels does not match given image size. Expected {} ({} * {}), got {}. ",
+            expected_pixel_count,
+            size.0,
+            size.1,
+            rgba_pixels.len()
+        )));
+    }
+    let config = construct_config(
+        colormode,
+        hierarchical,
+        mode,
+        filter_speckle,
+        color_precision,
+        layer_difference,
+        corner_threshold,
+        length_threshold,
+        max_iterations,
+        splice_threshold,
+        path_precision,
+    );
+    let mut flat_pixels: Vec<u8> = vec![];
+    for (r, g, b, a) in rgba_pixels {
+        flat_pixels.push(r);
+        flat_pixels.push(g);
+        flat_pixels.push(b);
+        flat_pixels.push(a);
+    }
+    let mut img = ColorImage::new();
+    img.pixels = flat_pixels;
+    (img.width, img.height) = size;
+
+    let svg =
+        convert(img, config).map_err(|_| PyException::new_err("Failed to convert the image. "))?;
+    Ok(format!("{}", svg))
+}
+
+fn construct_config(
+    colormode: Option<&str>,
+    hierarchical: Option<&str>,
+    mode: Option<&str>,
+    filter_speckle: Option<usize>,
+    color_precision: Option<i32>,
+    layer_difference: Option<i32>,
+    corner_threshold: Option<i32>,
+    length_threshold: Option<f64>,
+    max_iterations: Option<usize>,
+    splice_threshold: Option<i32>,
+    path_precision: Option<u32>,
+) -> Config {
     // TODO: enforce color mode with an enum so that we only
     // accept the strings 'color' or 'binary'
     let color_mode = match colormode.unwrap_or("color") {
@@ -52,7 +196,7 @@ fn convert_image_to_svg_py(
     let splice_threshold = splice_threshold.unwrap_or(45);
     let max_iterations = max_iterations.unwrap_or(10);
 
-    let config = Config {
+    Config {
         color_mode,
         hierarchical,
         filter_speckle,
@@ -65,15 +209,14 @@ fn convert_image_to_svg_py(
         splice_threshold,
         path_precision,
         ..Default::default()
-    };
+    }
-
-    convert_image_to_svg(&input_path, &output_path, config).unwrap();
-    Ok(())
 }
 
 /// A Python module implemented in Rust.
 #[pymodule]
 fn vtracer(_py: Python, m: &PyModule) -> PyResult<()> {
     m.add_function(wrap_pyfunction!(convert_image_to_svg_py, m)?)?;
+    m.add_function(wrap_pyfunction!(convert_raw_image_to_svg, m)?)?;
+    m.add_function(wrap_pyfunction!(convert_pixels_to_svg, m)?)?;
     Ok(())
 }

cmdapp/src/svg.rs

@@ -1,6 +1,7 @@
 use std::fmt;
 use visioncortex::{Color, CompoundPath, PointF64};
 
+#[derive(Debug, Clone)]
 pub struct SvgFile {
     pub paths: Vec<SvgPath>,
     pub width: usize,
@@ -8,6 +9,7 @@ pub struct SvgFile {
     pub path_precision: Option<u32>,
 }
 
+#[derive(Debug, Clone)]
 pub struct SvgPath {
     pub path: CompoundPath,
     pub color: Color,

cmdapp/vtracer/README.md

@@ -52,7 +52,17 @@ vtracer.convert_image_to_svg_py(inp, out)
 # Single-color example. Good for line art, and much faster than full color:
 vtracer.convert_image_to_svg_py(inp, out, colormode='binary')
 
-# All the bells & whistles
+# Convert from raw image bytes
+input_img_bytes: bytes = get_bytes() # e.g. reading bytes from a file or a HTTP request body
+svg_str: str = vtracer.convert_raw_image_to_svg(input_img_bytes, img_format='jpg')
+
+# Convert from RGBA image pixels
+from PIL import Image
+img = Image.open(input_path).convert('RGBA')
+pixels: list[tuple[int, int, int, int]] = list(img.getdata())
+svg_str: str = vtracer.convert_pixels_to_svg(pixels, img.size)
+
+# All the bells & whistles, also applicable to convert_raw_image_to_svg and convert_pixels_to_svg. 
 vtracer.convert_image_to_svg_py(inp,
                                 out,
                                 colormode = 'color',        # ["color"] or "binary"

cmdapp/vtracer/__init__.py

@@ -1 +1,2 @@
-from .vtracer import convert_image_to_svg_py
+from .vtracer import (convert_image_to_svg_py, convert_pixels_to_svg,
+                      convert_raw_image_to_svg)

cmdapp/vtracer/vtracer.pyi

@@ -15,3 +15,35 @@ def convert_image_to_svg_py(image_path: str,
                             path_precision: Optional[int] = None,   # default: 8
                         ) -> None:
     ...
+
+def convert_raw_image_to_svg(img_bytes: bytes,
+                            img_format: Optional[str] = None,       # Format of the image (e.g. 'jpg', 'png'... A full list of supported formats can be found [here](https://docs.rs/image/latest/image/enum.ImageFormat.html)). If not provided, the image format will be guessed based on its contents. 
+                            colormode: Optional[str] = None,        # ["color"] or "binary"
+                            hierarchical: Optional[str] = None,     # ["stacked"] or "cutout"
+                            mode: Optional[str] = None,             # ["spline"], "polygon", "none"
+                            filter_speckle: Optional[int] = None,   # default: 4
+                            color_precision: Optional[int] = None,  # default: 6
+                            layer_difference: Optional[int] = None, # default: 16
+                            corner_threshold: Optional[int] = None, # default: 60   
+                            length_threshold: Optional[float] = None, # in [3.5, 10] default: 4.0
+                            max_iterations: Optional[int] = None,   # default: 10
+                            splice_threshold: Optional[int] = None, # default: 45
+                            path_precision: Optional[int] = None,   # default: 8
+                        ) -> str:
+    ...
+
+def convert_pixels_to_svg(rgba_pixels: list[tuple[int, int, int, int]],
+                            size: tuple[int, int],
+                            colormode: Optional[str] = None,        # ["color"] or "binary"
+                            hierarchical: Optional[str] = None,     # ["stacked"] or "cutout"
+                            mode: Optional[str] = None,             # ["spline"], "polygon", "none"
+                            filter_speckle: Optional[int] = None,   # default: 4
+                            color_precision: Optional[int] = None,  # default: 6
+                            layer_difference: Optional[int] = None, # default: 16
+                            corner_threshold: Optional[int] = None, # default: 60   
+                            length_threshold: Optional[float] = None, # in [3.5, 10] default: 4.0
+                            max_iterations: Optional[int] = None,   # default: 10
+                            splice_threshold: Optional[int] = None, # default: 45
+                            path_precision: Optional[int] = None,   # default: 8
+                        ) -> str:
+    ...

webapp/Cargo.toml

@@ -22,7 +22,7 @@ console_log = { version = "0.2", features = ["color"] }
 wasm-bindgen = { version = "0.2", features = ["serde-serialize"]  }
 serde = { version = "1.0", features = ["derive"] }
 serde_json = "1.0"
-visioncortex = "0.6.0"
+visioncortex = "0.8.1"
 
 # The `console_error_panic_hook` crate provides better debugging of panics by
 # logging them with `console.error`. This is great for development, but requires

webapp/app/index.js

@@ -35,7 +35,11 @@ document.addEventListener('paste', function (e) {
 
 // Download as SVG
 document.getElementById('export').addEventListener('click', function (e) {
-    const blob = new Blob([new XMLSerializer().serializeToString(svg)], {type: 'octet/stream'}),
+    const blob = new Blob([
+        `<?xml version="1.0" encoding="UTF-8"?>\n`,
+        `<!-- Generator: visioncortex VTracer -->\n`,
+        new XMLSerializer().serializeToString(svg)
+    ], {type: 'octet/stream'}),
     url = window.URL.createObjectURL(blob);
 
     this.href = url;
@@ -444,7 +448,7 @@ class ConverterRunner {
         this.converter.init();
         this.stopped = false;
         if (clustering_mode == 'binary') {
-            svg.style.background = '#000';
+            svg.style.background = '#fff';
             canvas.style.display = 'none';
         } else {
             svg.style.background = '';

webapp/src/conversion/binary_image.rs

@@ -77,7 +77,7 @@ impl BinaryImageConverter {
                     self.params.max_iterations,
                     self.params.splice_threshold
                 );
-                let color = Color::color(&ColorName::White);
+                let color = Color::color(&ColorName::Black);
                 self.svg.prepend_path(
                     &paths,
                     &color,

webapp/src/conversion/color_image.rs

@@ -1,6 +1,6 @@
 use wasm_bindgen::prelude::*;
-use visioncortex::PathSimplifyMode;
+use visioncortex::{Color, ColorImage, PathSimplifyMode};
-use visioncortex::color_clusters::{IncrementalBuilder, Clusters, Runner, RunnerConfig, HIERARCHICAL_MAX};
+use visioncortex::color_clusters::{Clusters, Runner, RunnerConfig, HIERARCHICAL_MAX, IncrementalBuilder, KeyingAction};
 
 use crate::canvas::*;
 use crate::svg::*;
@@ -8,6 +8,8 @@ use crate::svg::*;
 use serde::Deserialize;
 use super::util;
 
+const KEYING_THRESHOLD: f32 = 0.2;
+
 #[derive(Debug, Deserialize)]
 pub struct ColorImageConverterParams {
     pub canvas_id: String,
@@ -67,7 +69,26 @@ impl ColorImageConverter {
     pub fn init(&mut self) {
         let width = self.canvas.width() as u32;
         let height = self.canvas.height() as u32;
-        let image = self.canvas.get_image_data_as_color_image(0, 0, width, height);
+        let mut image = self.canvas.get_image_data_as_color_image(0, 0, width, height);
+
+        let key_color = if Self::should_key_image(&image) {
+            if let Ok(key_color) = Self::find_unused_color_in_image(&image) {
+                for y in 0..height as usize {
+                    for x in 0..width as usize {
+                        if image.get_pixel(x, y).a == 0 {
+                            image.set_pixel(x, y, &key_color);
+                        }
+                    }
+                }
+                key_color
+            } else {
+                Color::default()
+            }
+        } else {
+            // The default color is all zeroes, which is treated by visioncortex as a special value meaning no keying will be applied.
+            Color::default()
+        };
+
         let runner = Runner::new(RunnerConfig {
             diagonal: self.params.layer_difference == 0,
             hierarchical: HIERARCHICAL_MAX,
@@ -78,6 +99,12 @@ impl ColorImageConverter {
             is_same_color_b: 1,
             deepen_diff: self.params.layer_difference,
             hollow_neighbours: 1,
+            key_color,
+            keying_action: if self.params.hierarchical == "cutout" {
+                KeyingAction::Keep
+            } else {
+                KeyingAction::Discard
+            },
         }, image);
         self.stage = Stage::Clustering(runner.start());
     }
@@ -108,6 +135,8 @@ impl ColorImageConverter {
                                 is_same_color_b: 1,
                                 deepen_diff: 0,
                                 hollow_neighbours: 0,
+                                key_color: Default::default(),
+                                keying_action: KeyingAction::Discard,
                             }, image);
                             self.stage = Stage::Reclustering(runner.start());
                         },
@@ -167,4 +196,56 @@ impl ColorImageConverter {
         }) as i32
     }
 
+    fn color_exists_in_image(img: &ColorImage, color: Color) -> bool {
+        for y in 0..img.height {
+            for x in 0..img.width {
+                let pixel_color = img.get_pixel(x, y);
+                if pixel_color.r == color.r && pixel_color.g == color.g && pixel_color.b == color.b {
+                    return true
+                }
+            }
+        }
+        false
+    }
+    
+    fn find_unused_color_in_image(img: &ColorImage) -> Result<Color, String> {
+        let special_colors = IntoIterator::into_iter([
+            Color::new(255, 0,   0),
+            Color::new(0,   255, 0),
+            Color::new(0,   0,   255),
+            Color::new(255, 255, 0),
+            Color::new(0,   255, 255),
+            Color::new(255, 0,   255),
+            Color::new(128, 128, 128),
+        ]);
+        for color in special_colors {
+            if !Self::color_exists_in_image(img, color) {
+                return Ok(color);
+            }
+        }
+        Err(String::from("unable to find unused color in image to use as key"))
+    }
+    
+    fn should_key_image(img: &ColorImage) -> bool {
+        if img.width == 0 || img.height == 0 {
+            return false;
+        }
+    
+        // Check for transparency at several scanlines
+        let threshold = ((img.width * 2) as f32 * KEYING_THRESHOLD) as usize;
+        let mut num_transparent_pixels = 0;
+        let y_positions = [0, img.height / 4, img.height / 2, 3 * img.height / 4, img.height - 1];
+        for y in y_positions {
+            for x in 0..img.width {
+                if img.get_pixel(x, y).a == 0 {
+                    num_transparent_pixels += 1;
+                }
+                if num_transparent_pixels >= threshold {
+                    return true;
+                }
+            }
+        }
+    
+        false
+    }
 }

webapp/src/conversion/mod.rs

@@ -1,6 +1,3 @@
 mod binary_image;
 mod color_image;
 mod util;
-
-pub use binary_image::*;
-pub use color_image::*;