Change .to_image(...)'s approach ...

... preferring to composite with a white background instead of removing
the alpha channel. This seems to more reliably produce high-quality
conversions (easier to read, fewer conversion artifacts) than either the
prior or other previous approaches.

pdfplumber/display.py

@@ -53,12 +53,12 @@ def postprocess(img: WandImage) -> WandImage:
 
     with WandImage(resolution=resolution, filename=filename, file=file) as img_init:
         img = postprocess(img_init)
-        if img.alpha_channel:
-            img.background_color = WandColor("white")
-            img.alpha_channel = "remove"
-        with img.convert("png") as png:
-            im = PIL.Image.open(BytesIO(png.make_blob()))
-            return im.convert("RGB")
+        with WandImage(
+            width=img.width, height=img.height, background=WandColor("white")
+        ) as bg:
+            bg.composite(img, 0, 0)
+            im = PIL.Image.open(BytesIO(bg.make_blob("png")))
+        return im.convert("RGBA")
 
 
 class PageImage:

tests/test_display.py

@@ -78,7 +78,4 @@ def test_outline_chars(self):
     def test__repr_png_(self):
         png = self.im._repr_png_()
         assert isinstance(png, bytes)
-        assert len(png) in (
-            71948,
-            61247,
-        )  # PNG encoder seems to work differently on different setups
+        assert len(png) == 79774