How to make thumbnails using Imagick::resizeImage

Question

I am using Imagick::resizeImage to create a thumbnail PNG image of each page of a pdf file. However, the image I am getting back is really blurry. How can I make it a little sharper? Any pointers would be really appreciated.

I have tried adjusting the 'blur' paramter of Imagick::resizeImage between 0.1 - 1, without success.

$pdfPage = '1Mpublic.pdf[0]';
$im = new imagick($pdfPage);
$im->setImageFormat('png');

// Resize thumbnail image
$imgHeight =  $im -> getImageHeight();
$imgWidth = $im -> getImageWidth();
$desiredWidth = 200;
$desiredHeight = resizedImageHeight($imgWidth, $imgHeight, $desiredWidth);

$im -> setResolution(1500, 1500);
$im -> resizeImage($desiredWidth, $desiredHeight, imagick::STYLE_NORMAL, 0.1);

 /* Resize image */
function resizedImageHeight($imgWidth, $imgHeight, $desiredImgWidth){

    $quoient = $imgWidth/$imgHeight;
    $height = $desiredImgWidth/$quoient;

    return $height;
}

original pdf link:

https://www.ncbi.nlm.nih.gov/pmc/articles/PMC4905263/pdf/ksgt-06-04-1091539.pdf

Answer 1

Unfortunately, I do not know Imagick that well. But in Imagemagick command line, I would do what is sometimes called supersampling. That is use a large density to read the PDF, then resize down by the inverse scale factor.

For example, nominal density is 72 dpi. I would read the input PDF at 4*72=288 dpi. Then after rasterizing, I would resize by 1/4=25% or for a larger result by something larger than 25%, say 50%. Here is your first page done both ways:

convert -density 288 ksgt-06-04-1091539.pdf[0] -resize 25% result1.png

convert -density 288 ksgt-06-04-1091539.pdf[0] -resize 50% result2.png

In Imagick, the key is something like:

$imagick = new Imagick(); 
$imagick->setImageResolution(288, 288);
$imagick->readImage('myfile.pdf'); 

Answer 2

Rather than rendering and then resizing the raster, it might be better to render the PDF to the right number of pixels in the first place. It'll be faster, and you can be sure that the amount of sharpness is correct for the content.

For example:

$ time convert -density 50 ksgt-06-04-1091539.pdf[0] x2.png
real    0m0.325s
user    0m0.299s
sys 0m0.024s

Makes:

The -density 50 makes a page about the same number of pixels across as your sample, 425.

In imagick you could do it like this (as @fmw42's excellent answer already says):

#!/usr/bin/env php
<?php

$target_width = 400;

# get natural size, calculate density we need for target width
$im = new imagick();
$im->pingImage($argv[1]);
$geo = $im->getImageGeometry();
$natural_width = $geo['width'];
$density = 72.0 * $target_width / $natural_width;

# load at correct resolution for target_width
$im = new imagick();
$im->setResolution($density, $density);
$im->readImage($argv[1]);

# and write back
$im->writeImage($argv[2]);

Doing both the ping and the read is a little slow in imagick, unfortunately:

$ time ./pdfthumb.php ksgt-06-04-1091539.pdf x.png
real    0m2.773s
user    0m2.737s
sys 0m0.036s

It's not imagick, but vipsthumbnail can do the ping and read in one operation:

$ time vipsthumbnail ksgt-06-04-1091539.pdf -s 400x -o x2.png
real    0m0.064s
user    0m0.064s
sys 0m0.011s

It might be worth considering if speed is important. libvips has a php binding so you can call it directly, but if you do that you'll run into awful licensing problems because it uses the GPL library poppler for PDF rendering, sigh. ImageMagick uses GhostScript and shells out to that for the same reason.