利用k-means聚类算法识别图片主色调

发布:smiling　来源: PHP粉丝网　添加日期：2014-08-27 14:03:34　浏览: 评论：0　

由于使用php来写图片主色调识别功能太麻烦了,所以我给大家介绍利用利用k-means聚类算法识别图片主色调方法.

识别图片主色调这个,网上貌似有几种方法,不过最准确,最优雅的解决方案还是利用聚类算法来做.

直接上代码,不过,我测试结果表示,用PHP来做,效率不佳,PHP不适合做这种大规模运算,用nodejs做效率可以高出100倍左右,代码如下:

<?php

$start = microtime(TRUE);

main();

function main($img = ‘colors_files/T1OX3eXldXXXcqfYM._111424.jpg’)

{



list($width, $height, $mime_code) = getimagesize($img);

//开源代码phpfensi.com

$im = null;

$point = array();

switch ($mime_code)

{

# jpg

case 2:

$im =imagecreatefromjpeg($img);

break;

# png

case 3:



default:

exit(‘擦，什么图像？解析不了啊’);

}



$new_width = 100;

$new_height = 100;

$pixel = imagecreatetruecolor($new_width, $new_height);

imagecopyresampled($pixel, $im, 0, 0, 0, 0, $new_width, $new_height, $width, $height);



run_time();



$i = $new_width;

while ($i–)

{

# reset高度

$k = $new_height;

while ($k–)

{

$rgb = ImageColorAt($im, $i, $k);

array_push($point, array(‘r’=>($rgb >> 16) & 0xFF, ‘g’=>($rgb >> 8) & 0xFF, ‘b’=>$rgb & 0xFF));

}

}

imagedestroy($im);

imagedestroy($pixel);

run_time();



$color = kmeans($point);



run_time();



foreach ($color as $key => $value)

&nb

sp; {

echo ‘<br><span style=“background-color:’ . RGBToHex($value[0]) . ‘” >’ . RGBToHex($value[0]) . ‘</span>’;

}



}



function run_time()

{

global $start;

echo ‘<br/>消耗:’, microtime(TRUE) – $start;

}



function kmeans($point=array(), $k=3, $min_diff=1)

{

global $ii;

$point_len = count($point);

$clusters = array();

$cache = array();



for ($i=0; $i < 256; $i++)

{

$cache[$i] = $i*$i;

}

# 随机生成k值

$i = $k;

$index = 0;

while ($i–)

{

$index = mt_rand(1,$point_len-100);

array_push($clusters, array($point[$index], array($point[$index])));

}

run_time();

$point_list = array();



$run_num = 0;



while (TRUE)

{

foreach ($point as $value)

{

$smallest_distance = 10000000;



# 求出距离最小的点

# index用于保存point最靠近的k值

$index = 0;

$i = $k;

while ($i–)

{

$distance = 0;

foreach ($value as $key => $p1)

{

&n

bsp; if ($p1 > $clusters[$i][0][$key])

{

$distance += $cache[$p1 - $clusters[$i][0][$key]];

}

else

{

$distance += $cache[$clusters[$i][0][$key] – $p1];

}

}

$ii++;

if ($distance < $smallest_distance)

{

$smallest_distance = $distance;

$index = $i;

}

}

$point_list[$index][] = $value;

}



$diff = 0;

# 1个1个迭代k值

$i = $k;

while ($i–)

{

$old = $clusters[$i];



# 移到到队列中心

$center = calculateCenter($point_list[$i], 3);

# 形成新的k值集合队列

$new_cluster = array($center, $point_list[$i]);

$clusters[$i] = $new_cluster;



# 计算新的k值与队列所在点的位置

$diff = euclidean($old[0], $center);

}

# 判断是否已足够聚合

if ($diff < $min_diff)

{

break;

>

}

}

echo ‘—>’.$ii;



return $clusters;

}



# 计算2点距离

$ii = 0;

function euclidean($p1, $p2)

{



$s = 0;

foreach ($p1 as $key => $value)

{



$temp = ($value – $p2[$key]);

$s += $temp*$temp;

}

return sqrt($s);

}



# 移动k值到所有点的中心

function calculateCenter($point_list, $attr_num) {

$vals = array();

$point_num = 0;

$keys = array_keys($point_list[0]);

foreach($keys as $value)

{

$vals[$value] = 0;

}

foreach ($point_list as $arr)

{

$point_num++;

foreach ($arr as $key => $value)

{

$vals[$key] += $value;

}

}

foreach ($keys as $index)

{

$vals[$index] = $vals[$index] / $point_num;

}



return $vals;

}

function RGBToHex($r, $g=”, $b=”)

{

if (is_array($r))

{

$b = $r['b'];

$g = $r['g'];

$r = $r['r'];

}



$hex = “#”;

$hex.= str_pad(dechex($r), 2, ’0′, STR_PAD_LEFT);

$hex.= str_pad(dechex($g), 2, ’0′, STR_PAD_LEFT);

$hex.= str_pad(dechex($b), 2, ’0′, STR_PAD_LEFT);



return $hex;

}

?>