diff --git a/modules/extract/src/main/scala/docspell/extract/pdfbox/PdfboxPreview.scala b/modules/extract/src/main/scala/docspell/extract/pdfbox/PdfboxPreview.scala
index f1a273ac..b0ddadc7 100644
--- a/modules/extract/src/main/scala/docspell/extract/pdfbox/PdfboxPreview.scala
+++ b/modules/extract/src/main/scala/docspell/extract/pdfbox/PdfboxPreview.scala
@@ -6,10 +6,13 @@
 
 package docspell.extract.pdfbox
 
+import java.awt.RenderingHints
 import java.awt.image.BufferedImage
 import java.awt.image.RenderedImage
 import javax.imageio.ImageIO
 
+import scala.jdk.CollectionConverters._
+
 import cats.effect._
 import cats.implicits._
 import fs2.Chunk
@@ -17,6 +20,8 @@ import fs2.Stream
 
 import org.apache.commons.io.output.ByteArrayOutputStream
 import org.apache.pdfbox.pdmodel.PDDocument
+import org.apache.pdfbox.pdmodel.PDPage
+import org.apache.pdfbox.pdmodel.graphics.image.PDImageXObject
 import org.apache.pdfbox.rendering.PDFRenderer
 import scodec.bits.ByteVector
 
@@ -27,6 +32,7 @@ trait PdfboxPreview[F[_]] {
 }
 
 object PdfboxPreview {
+  private[this] val logger = org.log4s.getLogger
 
   def apply[F[_]: Sync](cfg: PreviewConfig): F[PdfboxPreview[F]] =
     Sync[F].pure(new PdfboxPreview[F] {
@@ -50,14 +56,52 @@ object PdfboxPreview {
     if (count <= 0 || page < 0 || count <= page) None
     else {
       val renderer = new PDFRenderer(pdoc)
+      renderer.setImageDownscalingOptimizationThreshold(0.85f)
+      val hints = new RenderingHints(
+        RenderingHints.KEY_RENDERING,
+        RenderingHints.VALUE_RENDER_QUALITY
+      )
+      hints.put(
+        RenderingHints.KEY_COLOR_RENDERING,
+        RenderingHints.VALUE_COLOR_RENDER_QUALITY
+      )
+      hints.put(
+        RenderingHints.KEY_INTERPOLATION,
+        RenderingHints.VALUE_INTERPOLATION_BICUBIC
+      )
+      hints.put(
+        RenderingHints.KEY_TEXT_ANTIALIASING,
+        RenderingHints.VALUE_TEXT_ANTIALIAS_ON
+      )
+      hints.put(RenderingHints.KEY_ANTIALIASING, RenderingHints.VALUE_ANTIALIAS_ON)
+      renderer.setRenderingHints(hints)
+      renderer.setSubsamplingAllowed(enableSubsampling(pdoc.getPage(page)))
+      logger.debug(s"Converting pdf page $page to image at dpi $dpi")
       Option(renderer.renderImageWithDPI(page, dpi))
     }
   }
 
+  private def enableSubsampling(page: PDPage): Boolean = {
+    val res = page.getResources()
+    val largestImage =
+      res
+        .getXObjectNames()
+        .asScala
+        .map(name => res.getXObject(name))
+        .collect { case xobj: PDImageXObject => xobj }
+        .map { imgobj =>
+          val w = imgobj.getWidth()
+          val h = imgobj.getHeight()
+          logger.debug(s"Found image in pdf of size ${w}x${h} (${w * h}px)")
+          w * h
+        }
+        .maxOption
+    largestImage.exists(_ > 10 * 1024 * 1024)
+  }
+
   private def pngStream[F[_]](img: RenderedImage): Stream[F, Byte] = {
     val out = new ByteArrayOutputStream()
     ImageIO.write(img, "PNG", out)
     Stream.chunk(Chunk.byteVector(ByteVector.view(out.toByteArray())))
   }
-
 }