diff --git a/utils/README.md b/utils/README.md
new file mode 100644
index 0000000..8c56a9b
--- /dev/null
+++ b/utils/README.md
@@ -0,0 +1,12 @@
+## Utilities
+
+### decode_cloanto_rom.py
+
+Turns a Cloanto ROM into a regular Amiga ROM image. Useful e.g. when you have
+an emulator/machine that can't read Cloanto ROMs directly. WinUAE has this
+built in.
+
+### png2image.py
+
+Converts a PNG image into a AmigaOS Image structure that can be used directly
+in Intuition programs.
diff --git a/utils/png2image.py b/utils/png2image.py
new file mode 100755
index 0000000..1de58a0
--- /dev/null
+++ b/utils/png2image.py
@@ -0,0 +1,75 @@
+#!/usr/bin/python
+
+"""
+For license, see gpl-3.0.txt
+"""
+from PIL import Image
+import argparse
+import math
+import sys
+
+def chunks(l, n):
+    for i in xrange(0, len(l), n):
+        yield l[i:i+n]
+
+def color_to_plane_bits(color, depth):
+    """returns the bits for a given pixel in a list, lowest to highest plane"""
+    result = [0] * depth
+    for bit in range(depth):
+        if color & (1 << bit) != 0:
+            result[bit] = 1
+    return result
+
+def write_amiga_image(image, outfile):
+    imdata = im.getdata()
+    width, height = im.size
+    colors = [i for i in chunks(map(ord, im.palette.tobytes()), 3)]
+    depth = int(math.log(len(colors), 2))
+
+    map_words_per_row = width / 16
+    if width % 16 > 0:
+        map_words_per_row += 1
+
+    # create the converted planar data
+    planes = [[0] * (map_words_per_row * height) for _ in range(depth)]
+    for y in range(height):
+        x = 0
+        while x < width:
+            # build a word for each plane
+            for i in range(min(16, width - x)):
+                # get the palette index for pixel (x + i, y)
+                color = imdata[y * width + x + i]  # color index
+                planebits = color_to_plane_bits(color, depth)
+                # now we need to "or" the bits into the words in their respective planes
+                wordidx = (x + i) / 16  # word number in current row
+                pos = y * map_words_per_row + wordidx  # list index in the plane
+                for planeidx in range(depth):
+                    if planebits[planeidx]:
+                        planes[planeidx][pos] |= (1 << (15 - (x + i) % 16)) # 1 << ((x + i) % 16)
+            x += 16
+
+    outfile.write('#include <intuition/intuition.h>\n\n')
+
+    outfile.write("/* Ensure that this data is within chip memory or you'll see nothing !!! */\n");
+    outfile.write('UWORD imdata[] = {\n')
+    for plane in planes:
+        for chunk in chunks(plane, map_words_per_row):
+            outfile.write(','.join(map(str, chunk)))
+            outfile.write(',\n')
+        outfile.write('\n')
+    outfile.write('};\n\n')
+
+
+    planepick = 2 ** depth - 1  # currently we always use all of the planes
+    outfile.write('struct Image image = {\n')
+    outfile.write('  0, 0, %d, %d, %d, imdata,\n' % (width, height, depth));
+    outfile.write('  %d, 0, NULL\n' % planepick)  # PlanePick, PlaneOnOff
+    outfile.write('};\n')
+
+if __name__ == '__main__':
+    parser = argparse.ArgumentParser(description='Amiga Image converter')
+    parser.add_argument('pngfile')
+
+    args = parser.parse_args()
+    im = Image.open(args.pngfile)
+    write_amiga_image(im, sys.stdout)