* @copyright 2021 Smiley * @license Apache-2.0 */ namespace chillerlan\QRCode\Detector; use Exception, RuntimeException; use chillerlan\QRCode\Decoder\BitMatrix; use function array_fill, count, sprintf; /** * Implementations of this class can, given locations of finder patterns for a QR code in an * image, sample the right points in the image to reconstruct the QR code, accounting for * perspective distortion. It is abstracted since it is relatively expensive and should be allowed * to take advantage of platform-specific optimized implementations, like Sun's Java Advanced * Imaging library, but which may not be available in other environments such as J2ME, and vice * versa. * * The implementation used can be controlled by calling {@link #setGridSampler(GridSampler)} * with an instance of a class which implements this interface. * * @author Sean Owen */ final class GridSampler{ /** *

Checks a set of points that have been transformed to sample points on an image against * the image's dimensions to see if the point are even within the image.

* *

This method will actually "nudge" the endpoints back onto the image if they are found to be * barely (less than 1 pixel) off the image. This accounts for imperfect detection of finder * patterns in an image where the QR Code runs all the way to the image border.

* *

For efficiency, the method will check points from either end of the line until one is found * to be within the image. Because the set of points are assumed to be linear, this is valid.

* * @param \chillerlan\QRCode\Decoder\BitMatrix $bitMatrix image into which the points should map * @param float[] $points actual points in x1,y1,...,xn,yn form * * @throws \RuntimeException if an endpoint is lies outside the image boundaries */ private function checkAndNudgePoints(BitMatrix $bitMatrix, array $points):void{ $dimension = $bitMatrix->getDimension(); $nudged = true; $max = count($points); // Check and nudge points from start until we see some that are OK: for($offset = 0; $offset < $max && $nudged; $offset += 2){ $x = (int)$points[$offset]; $y = (int)$points[$offset + 1]; if($x < -1 || $x > $dimension || $y < -1 || $y > $dimension){ throw new RuntimeException(sprintf('checkAndNudgePoints 1, x: %s, y: %s, d: %s', $x, $y, $dimension)); } $nudged = false; if($x === -1){ $points[$offset] = 0.0; $nudged = true; } elseif($x === $dimension){ $points[$offset] = $dimension - 1; $nudged = true; } if($y === -1){ $points[$offset + 1] = 0.0; $nudged = true; } elseif($y === $dimension){ $points[$offset + 1] = $dimension - 1; $nudged = true; } } // Check and nudge points from end: $nudged = true; for($offset = count($points) - 2; $offset >= 0 && $nudged; $offset -= 2){ $x = (int)$points[$offset]; $y = (int)$points[$offset + 1]; if($x < -1 || $x > $dimension || $y < -1 || $y > $dimension){ throw new RuntimeException(sprintf('checkAndNudgePoints 2, x: %s, y: %s, d: %s', $x, $y, $dimension)); } $nudged = false; if($x === -1){ $points[$offset] = 0.0; $nudged = true; } elseif($x === $dimension){ $points[$offset] = $dimension - 1; $nudged = true; } if($y === -1){ $points[$offset + 1] = 0.0; $nudged = true; } elseif($y === $dimension){ $points[$offset + 1] = $dimension - 1; $nudged = true; } } } /** * Samples an image for a rectangular matrix of bits of the given dimension. The sampling * transformation is determined by the coordinates of 4 points, in the original and transformed * image space. * * @return \chillerlan\QRCode\Decoder\BitMatrix representing a grid of points sampled from the image within a region * defined by the "from" parameters * @throws \RuntimeException if image can't be sampled, for example, if the transformation defined * by the given points is invalid or results in sampling outside the image boundaries */ public function sampleGrid(BitMatrix $image, int $dimension, PerspectiveTransform $transform):BitMatrix{ if($dimension <= 0){ throw new RuntimeException('invalid matrix size'); } $bits = new BitMatrix($dimension); $points = array_fill(0, 2 * $dimension, 0.0); for($y = 0; $y < $dimension; $y++){ $max = count($points); $iValue = $y + 0.5; for($x = 0; $x < $max; $x += 2){ $points[$x] = ($x / 2) + 0.5; $points[$x + 1] = $iValue; } $transform->transformPoints($points); // Quick check to see if points transformed to something inside the image; // sufficient to check the endpoints $this->checkAndNudgePoints($image, $points); try{ for($x = 0; $x < $max; $x += 2){ if($image->get((int)$points[$x], (int)$points[$x + 1])){ // Black(-ish) pixel $bits->set($x / 2, $y); } } } catch(Exception $aioobe){//ArrayIndexOutOfBoundsException // This feels wrong, but, sometimes if the finder patterns are misidentified, the resulting // transform gets "twisted" such that it maps a straight line of points to a set of points // whose endpoints are in bounds, but others are not. There is probably some mathematical // way to detect this about the transformation that I don't know yet. // This results in an ugly runtime exception despite our clever checks above -- can't have // that. We could check each point's coordinates but that feels duplicative. We settle for // catching and wrapping ArrayIndexOutOfBoundsException. throw new RuntimeException('ArrayIndexOutOfBoundsException'); } } return $bits; } }