pihash 0.5.0

A simple library for generating perceptual hashes for images and comparing images based on their perceptual hashes.
Documentation 
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
208
209
210
211
212
213
214
215
216
217
218
219
220
221
222
223
224
225
226
227
228
229
230
231
232
233
234
235
236

// Copyright 2016 Drew Short <drew@sothr.com>.
//
// Licensed under the MIT license<LICENSE-MIT or http://opensource.org/licenses/MIT>.
// This file may not be copied, modified, or distributed except according to those terms.
extern crate image;

use std::path::Path;

use cache::Cache;

use super::dft;
use super::dft::Transform;
use super::image::Pixel;
use super::prepare_image;
use super::{HashType, PerceptualHash, Precision, PreparedImage};

use self::image::{DynamicImage, GenericImage, GenericImageView};

pub struct PHash<'a> {
    prepared_image: Box<PreparedImage<'a>>,
}

impl<'a> PHash<'a> {
    pub fn new(path: &'a Path, precision: &Precision, cache: &Option<Cache>) -> Self {
        PHash {
            prepared_image: Box::new(prepare_image(&path, &HashType::PHash, &precision, cache)),
        }
    }
}

impl<'a> PerceptualHash for PHash<'a> {
    /**
     * Calculate the phash of the provided prepared image
     *
     * # Return
     *
     * Returns a u64 representing the value of the hash
     */
    fn get_hash(&self, cache: &Option<Cache>) -> u64 {
        match self.prepared_image.image {
            Some(ref image) => {
                // Get the image data into a vector to perform the DFT on.
                let (width, height) = image.dimensions();

                // Get 2d data to 2d FFT/DFT
                // Either from the cache or calculate it
                // Pretty fast already, so caching doesn't make a huge difference
                // Atleast compared to opening and processing the images
                let data_matrix: Vec<Vec<f64>> = match *cache {
                    Some(ref c) => {
                        match c.get_matrix_from_cache(
                            &Path::new(self.prepared_image.orig_path),
                            width as u32,
                        ) {
                            Some(matrix) => matrix,
                            None => {
                                let matrix = create_data_matrix(width, height, &image);
                                match c.put_matrix_in_cache(
                                    &Path::new(self.prepared_image.orig_path),
                                    width as u32,
                                    &matrix,
                                ) {
                                    Ok(_) => {}
                                    Err(e) => println!("Unable to store matrix in cache. {}", e),
                                };
                                matrix
                            }
                        }
                    }
                    None => create_data_matrix(width, height, &image),
                };

                // Only need the top left quadrant
                let target_width = (width / 4) as usize;
                let target_height = (height / 4) as usize;
                let dft_width = (width / 4) as f64;
                let dft_height = (height / 4) as f64;

                // Calculate the mean
                let mut total = 0f64;
                for x in 0..target_width {
                    for y in 0..target_height {
                        total += data_matrix[x][y];
                    }
                }
                let mean = total / (dft_width * dft_height);

                // Calculating a hash based on the mean
                let mut hash = 0u64;
                for x in 0..target_width {
                    for y in 0..target_height {
                        if data_matrix[x][y] >= mean {
                            hash |= 1;
                        } else {
                            hash |= 0;
                        }
                        hash <<= 1;
                    }
                }
                hash
            }
            None => 0u64,
        }
    }
}

fn create_data_matrix(width: u32, height: u32, image: &DynamicImage) -> Vec<Vec<f64>> {
    let mut data_matrix: Vec<Vec<f64>> = Vec::new();
    // Preparing the results
    for x in 0..width as usize {
        data_matrix.push(Vec::new());
        for y in 0..height {
            let pos_x = x as u32;
            let pos_y = y as u32;
            data_matrix[x].push(image.get_pixel(pos_x, pos_y).channels()[0] as f64);
        }
    }

    // Perform the 2D DFT operation on our matrix
    calculate_2d_dft(&mut data_matrix);
    data_matrix
}

// Use a 1D DFT to cacluate the 2D DFT.
//
// This is achieved by calculating the DFT for each row, then calculating the
// DFT for each column of DFT row data. This means that a 32x32 image with have
// 1024 1D DFT operations performed on it. (Slightly caclulation intensive)
//
// This operation is in place on the data in the provided vector
//
// Inspired by:
// http://www.inf.ufsc.br/~visao/khoros/html-dip/c5/s2/front-page.html
//
// Checked with:
// http://calculator.vhex.net/post/calculator-result/2d-discrete-fourier-transform
//
fn calculate_2d_dft(data_matrix: &mut Vec<Vec<f64>>) {
    // println!("{:?}", data_matrix);
    let width = data_matrix.len();
    let height = data_matrix[0].len();

    let mut complex_data_matrix = Vec::with_capacity(width);

    // Perform DCT on the columns of data
    for x in 0..width {
        let mut column: Vec<f64> = Vec::with_capacity(height);
        for y in 0..height {
            column.push(data_matrix[x][y]);
        }

        // Perform the DCT on this column
        // println!("column[{}] before: {:?}", x, column);
        let forward_plan = dft::Plan::new(dft::Operation::Forward, column.len());
        column.transform(&forward_plan);
        let complex_column = dft::unpack(&column);
        // println!("column[{}] after: {:?}", x, complex_column);
        complex_data_matrix.push(complex_column);
    }

    // Perform DCT on the rows of data
    for y in 0..height {
        let mut row = Vec::with_capacity(width);
        for x in 0..width {
            row.push(complex_data_matrix[x][y]);
        }
        // Perform DCT on the row
        // println!("row[{}] before: {:?}", y, row);
        let forward_plan = dft::Plan::new(dft::Operation::Forward, row.len());
        row.transform(&forward_plan);
        // println!("row[{}] after: {:?}", y, row);

        // Put the row values back
        for x in 0..width {
            data_matrix[x][y] = round_float(row[x].re);
        }
    }
}

fn round_float(f: f64) -> f64 {
    if f >= super::FLOAT_PRECISION_MAX_1 || f <= super::FLOAT_PRECISION_MIN_1 {
        f
    } else if f >= super::FLOAT_PRECISION_MAX_2 || f <= super::FLOAT_PRECISION_MIN_2 {
        (f * 10_f64).round() / 10_f64
    } else if f >= super::FLOAT_PRECISION_MAX_3 || f <= super::FLOAT_PRECISION_MIN_3 {
        (f * 100_f64).round() / 100_f64
    } else if f >= super::FLOAT_PRECISION_MAX_4 || f <= super::FLOAT_PRECISION_MIN_4 {
        (f * 1000_f64).round() / 1000_f64
    } else if f >= super::FLOAT_PRECISION_MAX_5 || f <= super::FLOAT_PRECISION_MIN_5 {
        (f * 10000_f64).round() / 10000_f64
    } else {
        (f * 100000_f64).round() / 100000_f64
    }
}

#[test]
fn test_2d_dft() {
    let mut test_matrix: Vec<Vec<f64>> = Vec::new();
    test_matrix.push(vec![1f64, 1f64, 1f64, 3f64]);
    test_matrix.push(vec![1f64, 2f64, 2f64, 1f64]);
    test_matrix.push(vec![1f64, 2f64, 2f64, 1f64]);
    test_matrix.push(vec![3f64, 1f64, 1f64, 1f64]);

    println!("{:?}", test_matrix[0]);
    println!("{:?}", test_matrix[1]);
    println!("{:?}", test_matrix[2]);
    println!("{:?}", test_matrix[3]);

    println!("Performing 2d DFT");
    calculate_2d_dft(&mut test_matrix);

    println!("{:?}", test_matrix[0]);
    println!("{:?}", test_matrix[1]);
    println!("{:?}", test_matrix[2]);
    println!("{:?}", test_matrix[3]);

    assert!(test_matrix[0][0] == 24_f64);
    assert!(test_matrix[0][1] == 0_f64);
    assert!(test_matrix[0][2] == 0_f64);
    assert!(test_matrix[0][3] == 0_f64);

    assert!(test_matrix[1][0] == 0_f64);
    assert!(test_matrix[1][1] == 0_f64);
    assert!(test_matrix[1][2] == -2_f64);
    assert!(test_matrix[1][3] == 2_f64);

    assert!(test_matrix[2][0] == 0_f64);
    assert!(test_matrix[2][1] == -2_f64);
    assert!(test_matrix[2][2] == -4_f64);
    assert!(test_matrix[2][3] == -2_f64);

    assert!(test_matrix[3][0] == 0_f64);
    assert!(test_matrix[3][1] == 2_f64);
    assert!(test_matrix[3][2] == -2_f64);
    assert!(test_matrix[3][3] == 0_f64);
}