imageconvolution

(PHP 5 >= 5.1.0, PHP 7, PHP 8)

imageconvolution用系数 div 和 offset 申请一个 3x3 的卷积矩阵

说明

imageconvolution ( resource $image , array $matrix , float $div , float $offset ) : bool

Applies a convolution matrix on the image, using the given coefficient and offset.

参数

image

由图象创建函数(例如imagecreatetruecolor())返回的图象资源。

matrix

A 3x3 matrix: an array of three arrays of three floats.

div

The divisor of the result of the convolution, used for normalization.

offset

Color offset.

返回值

成功时返回 true, 或者在失败时返回 false

范例

Example #1 Embossing the PHP.net logo

<?php
$image 
imagecreatefromgif('http://www.php.net/images/php.gif');

$emboss = array(array(200), array(0, -10), array(00, -1));
imageconvolution($image$emboss1127);

header('Content-Type: image/png');
imagepng($imagenull9);
?>

以上例程会输出:

Output of example : Embossing the PHP.net logo

Example #2 Gaussian blur

<?php
$image 
imagecreatetruecolor(180,40);

// Writes the text and apply a gaussian blur on the image
imagestring($image5108'Gaussian Blur Text'0x00ff00);
$gaussian = array(array(1.02.01.0), array(2.04.02.0), array(1.02.01.0));
imageconvolution($image$gaussian160);

// Rewrites the text for comparison
imagestring($image51018'Gaussian Blur Text'0x00ff00);

header('Content-Type: image/png');
imagepng($imagenull9);
?>

以上例程会输出:

Output of example : Gaussian blur

注释

参见

User Contributed Notes

magilvia 17-May-2011 01:36
Here's a good sharpen setting for thumbnail creation:

<?php
            $sharpenMatrix
= array
            (
                array(-
1.2, -1, -1.2),
                array(-
1, 20, -1),
                array(-
1.2, -1, -1.2)
            );

           
// calculate the sharpen divisor
           
$divisor = array_sum(array_map('array_sum', $sharpenMatrix));           

           
$offset = 0;
           
           
// apply the matrix
           
imageconvolution($img, $sharpenMatrix, $divisor, $offset);
?>
phunction.sf.net 15-May-2010 09:20
You don't need any custom function to calculate the $divisor of the $matrix, using array_map() and array_sum() does the trick:

<?php

$matrix
= array
(
    array(-
1, -1, -1),
    array(-
1, 16, -1),
    array(-
1, -1, -1),
);

$divisor = array_sum(array_map('array_sum', $matrix)); // 8

?>
fabien dot snauwaert at gmail dot com 07-Mar-2010 10:44
Normalization is necessary to keep the image balanced (else any filter may quickly turn the image almost completely black or white).

Here is a short, easy-to-use, class to handle normalization automatically and make for easier input of the 3x3 matrix:

The code respects the "array of three arrays" syntax for use with the imageconvolution() function and automatically calculates the necesarry divisor for normalization.

<?php

class ConvolutionFilter {
    public
$matrix;
    public
$div;
   
    public function
computeDiv() {
       
$this->div = array_sum ($this->matrix[0]) + array_sum ($this->matrix[1]) + array_sum ($this->matrix[2]);
    }

    function
__construct() {
       
$matrix = func_get_args();
       
$this->matrix = array(    array($matrix[0], $matrix[1], $matrix[2]),
                                array(
$matrix[3], $matrix[4], $matrix[5]),
                                array(
$matrix[6], $matrix[7], $matrix[8])
                                );
       
$this->computeDiv();
    }
}

?>

Example usage:

<?php

$gaussianFilter
=    new ConvolutionFilter(    1.0,    2.0,    1.0,
                                            
2.0,    3.0,    2.0,
                                            
1.0,    2.0,    1.0        );
imageconvolution($image, $gaussianFilter->matrix, $gaussianFilter->div, 0);

?>

Some common filters:

<?php

$identityFilter
=    new ConvolutionFilter(    0.0,    0.0,    0.0,
                                            
0.0,    1.0,    0.0,
                                            
0.0,    0.0,    0.0        );
$sharpenFilter =    new ConvolutionFilter(    0.0,    -1.0,    0.0,
                                             -
1.0,    5.0,    -1.0,
                                            
0.0,    -1.0,    0.0        );
$edgeFilter =        new ConvolutionFilter(    0.0,    1.0,    0.0,
                                            
1.0,    -4.0,    1.0,
                                            
0.0,    1.0,    0.0        );
$findEdgesFilter =    new ConvolutionFilter(    -1.0,    -1.0,    -1.0,
                                             -
2.0,    8.0,    -1.0,
                                             -
1.0,    -1.0,    -1.0        );

?>

Remember you can use imagefilter() for such basic needs but the above class will make it easier for you when you want to create your own filters.
Jase 13-Sep-2007 03:53
The comment below is an extremely good workaround

however, php did throw lots of warnings at me when i had error reporting set to E_ALL

this can be avoided with one line of code and no impact (as i can see) to the rest of the function

<?php
//include this file whenever you have to use imageconvolution...
//you can use in your project, but keep the comment below :)
//great for any image manipulation library
//Made by Chao Xu(Mgccl) 2/28/07
//www.webdevlogs.com
//V 1.0
if(!function_exists('imageconvolution')){
function
imageconvolution($src, $filter, $filter_div, $offset){
    if (
$src==NULL) {
        return
0;
    }
   
   
$sx = imagesx($src);
   
$sy = imagesy($src);
   
$srcback = ImageCreateTrueColor ($sx, $sy);
   
ImageCopy($srcback, $src,0,0,0,0,$sx,$sy);
   
    if(
$srcback==NULL){
        return
0;
    }
       
   
#FIX HERE
    #$pxl array was the problem so simply set it with very low values
   
$pxl = array(1,1);
   
#this little fix worked for me as the undefined array threw out errors

   
for ($y=0; $y<$sy; ++$y){
        for(
$x=0; $x<$sx; ++$x){
           
$new_r = $new_g = $new_b = 0;
           
$alpha = imagecolorat($srcback, $pxl[0], $pxl[1]);
           
$new_a = $alpha >> 24;
           
            for (
$j=0; $j<3; ++$j) {
               
$yv = min(max($y - 1 + $j, 0), $sy - 1);
                for (
$i=0; $i<3; ++$i) {
                       
$pxl = array(min(max($x - 1 + $i, 0), $sx - 1), $yv);
                   
$rgb = imagecolorat($srcback, $pxl[0], $pxl[1]);
                   
$new_r += (($rgb >> 16) & 0xFF) * $filter[$j][$i];
                   
$new_g += (($rgb >> 8) & 0xFF) * $filter[$j][$i];
                   
$new_b += ($rgb & 0xFF) * $filter[$j][$i];
                }
            }

           
$new_r = ($new_r/$filter_div)+$offset;
           
$new_g = ($new_g/$filter_div)+$offset;
           
$new_b = ($new_b/$filter_div)+$offset;

           
$new_r = ($new_r > 255)? 255 : (($new_r < 0)? 0:$new_r);
           
$new_g = ($new_g > 255)? 255 : (($new_g < 0)? 0:$new_g);
           
$new_b = ($new_b > 255)? 255 : (($new_b < 0)? 0:$new_b);

           
$new_pxl = ImageColorAllocateAlpha($src, (int)$new_r, (int)$new_g, (int)$new_b, $new_a);
            if (
$new_pxl == -1) {
               
$new_pxl = ImageColorClosestAlpha($src, (int)$new_r, (int)$new_g, (int)$new_b, $new_a);
            }
            if ((
$y >= 0) && ($y < $sy)) {
               
imagesetpixel($src, $x, $y, $new_pxl);
            }
        }
    }
   
imagedestroy($srcback);
    return
1;
}
}
?>
mgcclx at gmail dot com 28-Feb-2007 07:55
imageconvolution() does not appear in PHP with non-bundled GD libraries. It is a rare situation, but it still happens. That's why I wrote a replication of imageconvolution() in PHP. Compare to the post below, this one makes the use of offset and 30% faster.
Because it's written in PHP, it is 50 times slower than the bundled version.
Actually, this is a replication of gdimageconvolutaion() of GD library, it does not support data validating feature imageconvolution() have. But I guess people who uses this function knows their stuff.

THE SCRIPT:
<?php
//include this file whenever you have to use imageconvolution...
//you can use in your project, but keep the comment below :)
//great for any image manipulation library
//Made by Chao Xu(Mgccl) 2/28/07
//www.webdevlogs.com
//V 1.0
if(!function_exists('imageconvolution')){
function
imageconvolution($src, $filter, $filter_div, $offset){
    if (
$src==NULL) {
        return
0;
    }
   
   
$sx = imagesx($src);
   
$sy = imagesy($src);
   
$srcback = ImageCreateTrueColor ($sx, $sy);
   
ImageCopy($srcback, $src,0,0,0,0,$sx,$sy);
   
    if(
$srcback==NULL){
        return
0;
    }
       
    for (
$y=0; $y<$sy; ++$y){
        for(
$x=0; $x<$sx; ++$x){
           
$new_r = $new_g = $new_b = 0;
           
$alpha = imagecolorat($srcback, $pxl[0], $pxl[1]);
           
$new_a = $alpha >> 24;
           
            for (
$j=0; $j<3; ++$j) {
               
$yv = min(max($y - 1 + $j, 0), $sy - 1);
                for (
$i=0; $i<3; ++$i) {
                       
$pxl = array(min(max($x - 1 + $i, 0), $sx - 1), $yv);
                   
$rgb = imagecolorat($srcback, $pxl[0], $pxl[1]);
                   
$new_r += (($rgb >> 16) & 0xFF) * $filter[$j][$i];
                   
$new_g += (($rgb >> 8) & 0xFF) * $filter[$j][$i];
                   
$new_b += ($rgb & 0xFF) * $filter[$j][$i];
                }
            }

           
$new_r = ($new_r/$filter_div)+$offset;
           
$new_g = ($new_g/$filter_div)+$offset;
           
$new_b = ($new_b/$filter_div)+$offset;

           
$new_r = ($new_r > 255)? 255 : (($new_r < 0)? 0:$new_r);
           
$new_g = ($new_g > 255)? 255 : (($new_g < 0)? 0:$new_g);
           
$new_b = ($new_b > 255)? 255 : (($new_b < 0)? 0:$new_b);

           
$new_pxl = ImageColorAllocateAlpha($src, (int)$new_r, (int)$new_g, (int)$new_b, $new_a);
            if (
$new_pxl == -1) {
               
$new_pxl = ImageColorClosestAlpha($src, (int)$new_r, (int)$new_g, (int)$new_b, $new_a);
            }
            if ((
$y >= 0) && ($y < $sy)) {
               
imagesetpixel($src, $x, $y, $new_pxl);
            }
        }
    }
   
imagedestroy($srcback);
    return
1;
}
}
?>
interghost at crovortex dot com 31-Mar-2006 11:26
an implementation of this function for PHP versions <5.1
<?php
if(!function_exists("imageconvolution"))
{
 function
imageconvolution(&$img,$mat,$div,$off)
 {
   if(!
imageistruecolor($img) || !is_array($mat) || count($mat)!=3 || count($mat[0])!=3 || count($mat[1])!=3 || count($mat[2])!=3) return FALSE;
   unset(
$bojainfo);
   for(
$nx=0;$nx<imagesx($img)-1;$nx++)
   {
     for(
$ny=0;$ny<imagesy($img)-1;$ny++)
     {
       
$rgb=imagecolorat($img,$nx,$ny);
       
$bojainfo[$nx][$ny][r]=($rgb>>16)&0xFF;
       
$bojainfo[$nx][$ny][g]=($rgb>>8)&0xFF;
       
$bojainfo[$nx][$ny][b]=$rgb&0xFF;
     }
   }
   for(
$nx=1;$nx<imagesx($img)-1;$nx++)
   {
     for(
$ny=1;$ny<imagesy($img)-1;$ny++)
     {
       
$nr=$mat[0][0]*$bojainfo[$nx-1][$ny-1][r] + $mat[0][1]*$bojainfo[$nx][$ny-1][r] + $mat[0][2]*$bojainfo[$nx+1][$ny-1][r] + $mat[1][0]*$bojainfo[$nx-1][$ny][r] + $mat[1][1]*$bojainfo[$nx][$ny][r] + $mat[1][2]*$bojainfo[$nx+1][$ny][r] + $mat[2][0]*$bojainfo[$nx-1][$ny+1][r] + $mat[2][1]*$bojainfo[$nx][$ny+1][r] + $mat[2][2]*$bojainfo[$nx+1][$ny+1][r];
       
$nr=intval(round($nr/$div));
        if(
$nr<0) $nr=0;
        elseif(
$nr>255) $nr=255;
       
$ng=$mat[0][0]*$bojainfo[$nx-1][$ny-1][g]  + $mat[0][1]*$bojainfo[$nx][$ny-1][g] + $mat[0][2]*$bojainfo[$nx+1][$ny-1][g] + $mat[1][0]*$bojainfo[$nx-1][$ny][g] + $mat[1][1]*$bojainfo[$nx][$ny][g] + $mat[1][2]*$bojainfo[$nx+1][$ny][g] + $mat[2][0]*$bojainfo[$nx-1][$ny+1][g] + $mat[2][1]*$bojainfo[$nx][$ny+1][g] + $mat[2][2]*$bojainfo[$nx+1][$ny+1][g];
       
$ng=intval(round($ng/$div));
        if(
$ng<0) $ng=0;
        elseif(
$ng>255) $ng=255;
       
$nb=$mat[0][0]*$bojainfo[$nx-1][$ny-1][b] + $mat[0][1]*$bojainfo[$nx][$ny-1][b] + $mat[0][2]*$bojainfo[$nx+1][$ny-1][b] + $mat[1][0]*$bojainfo[$nx-1][$ny][b] + $mat[1][1]*$bojainfo[$nx][$ny][b] + $mat[1][2]*$bojainfo[$nx+1][$ny][b] + $mat[2][0]*$bojainfo[$nx-1][$ny+1][b] + $mat[2][1]*$bojainfo[$nx][$ny+1][b] + $mat[2][2]*$bojainfo[$nx+1][$ny+1][b];
       
$nb=intval(round($nb/$div));
        if(
$nb<0) $nb=0;
        elseif(
$nb>255) $nb=255;
       
$nrgb=($nr<<16)+($ng<<8)+$nb;
        if(!
imagesetpixel($img,$nx,$ny,$nrgb)) return FALSE;
     }
   }
   return
TRUE;
  }
}
?>

it's a bit slowish so I wouldn't recommend big images, also offset is not implemented (don't know what it's suppose to do)
mlconnor at yahoo dot com 21-Mar-2006 09:00
I've seen many people come up with ways to do a drop shadow behind a rectangle such as a picture.  I haven't found one yet that was fast, PHP 4 complaint, and nice looking.  Here is one I came up with last night.  It takes an image, fills it with the background, and creates a blurred drop shadow at the specified coords using the colors and the distance offset specified.  It looks great!!!

  function blurRect(&$image, $distance, $rectX1, $rectY1, $rectX2, $rectY2, $shadowR, $shadowG, $shadowB, $backR, $backG, $backB) {

    $potentialOverlap = ($distance * 2) * ($distance * 2);

    $backgroundColor = imagecolorallocate($image, $backR, $backG, $backB);
    $shadowColor = imagecolorallocate($image, $shadowR, $shadowG, $shadowB);

    $imageWidth = imagesx($image);
    $imageHeight = imagesy($image);

    imageFilledRectangle($image, 0, 0, $imageWidth - 1, $imageHeight - 1, $backgroundColor);
    imageFilledRectangle($image, $rectX1, $rectY1, $rectX2, $rectY2, $shadowColor);

    for ( $pointX = $rectX1 - $distance; $pointX < $imageWidth; $pointX++ ) {
      for ( $pointY = $rectY1 - $distance; $pointY < $imageHeight; $pointY++ ) {

        if ( $pointX > $rectX1 + $distance &&
             $pointX < $rectX2 - $distance &&
             $pointY > $rectY1 + $distance &&
             $pointY < $rectY2 - $distance ) {
          $pointY = $rectY2 - $distance;
        }

        $boxX1 = $pointX - $distance;
        $boxY1 = $pointY - $distance;
        $boxX2 = $pointX + $distance;
        $boxY2 = $pointY + $distance;

        $xOverlap = max(0, min($boxX2, $rectX2) - max($boxX1, $rectX1));
        $yOverlap = max(0, min($boxY2, $rectY2) - max($boxY1, $rectY1));

        $totalOverlap = $xOverlap * $yOverlap;
        $shadowPcnt = $totalOverlap / $potentialOverlap;
        $backPcnt = 1.0 - $shadowPcnt;

        $newR = $shadowR * $shadowPcnt + $backR * $backPcnt;
        $newG = $shadowG * $shadowPcnt + $backG * $backPcnt;
        $newB = $shadowB * $shadowPcnt + $backB * $backPcnt;

        $newcol = imagecolorallocate($image, $newR, $newG, $newB);
        imagesetpixel($image, $pointX, $pointY, $newcol);
      }
    }
  }
dyer85 at gmail dot com 07-Jan-2006 02:24
Took me a while, but thanks to a couple of the user notes on the array_values PHP documentation page, I was able to come up with a way to dynamically compute the divisor.

I'm using PHP 5.1.0b2 on Win32 with the bundled GD library. When I try and use the imageconvolution function, whether normally, or via the functions below, the resulting image (I've only tried JPEGs and GIFs), always comes out far too bright, even when the divisor makes matrix sum equal to 1. The only thing that would reduce the brightness was to make the offset argument ridiculously large. So, I'm not sure if this effects anyone else.

Here are the functions with an example:

<?php
$im
= imagecreatefromjpeg('path/to/pic.jpg');
$matrix = array(   array(5,5,5),
                   array(
5,15,5),
                   array(
5,5,5) );
makeFilter($im, $matrix);

header ( 'Content-Type: image/jpeg' );
imagejpeg($im);
imagedestroy($im);

/**
 * functions
 */
// This flattens the 3X3 array matrix, so we can get the sum of all the values
function array_flatten($array) {
    (array)
$tempArray = array();

    foreach (
$array as $value ) {
        if (
is_array($value) ) {
           
$tempArray = array_merge($tempArray, array_flatten($value));
        } else {
           
$tempArray[] = $value;
        }
    }

    return
$tempArray;
}

// Creates the divisor value dynamically, and passes offset
function makeFilter($resource, $matrix, $offset=1.0) {
    global $
$resource;
    (float)
$divisor = array_sum(array_flatten($matrix));
    return
imageconvolution($resource, $matrix, $divisor, $offset) ? true : false;
}
?>
26-Oct-2005 05:32
The example below didn't provide a 3x3 matrix. Correctly, it's a multidimensional array.

<?php
    $matrix
= array(    array( -1, -1, -1 ),
                        array( -
1, 16, -1 ),
                        array( -
1, -1, -1 ) );
?>
timeshifting at gmail dot com 24-Aug-2005 03:40
Matrices can be used for sharpening, blurring, edge detection, etc, ala Photoshop.

A sharpening example:

<?php

$sharpenMatrix
= array(-1,-1,-1,-1,16,-1,-1,-1,-1);
$divisor = 8;
$offset = 0;

imageconvolution($myImage, $sharpenMatrix, $divisor, $offset);

?>

Below is some information on building different kinds of matrices. (If you have photoshop (or PSP, GIMP) you can test out your matrices before applying them in PHP)

http://loriweb.pair.com/8udf-basics.html (covers blurs)
http://loriweb.pair.com/8udf-sharpen.html
http://loriweb.pair.com/8udf-edges.html
http://loriweb.pair.com/8udf-emboss.html