René's URL Explorer Experiment

{"@context":"https://schema.org","@type":"DiscussionForumPosting","headline":"[Bug]: Rounding to pixel boundary misaligns images in vector graphics outputs","articleBody":"### Bug summary\n\nWhen using Axes.imshow(), images are scaled to integer pixel size (`_ImageBase._make_image`). This makes images notably too large when using vector graphics output, misaligning them with tick marks and plotted lines.\n\n### Code for reproduction\n\n```python\nimport numpy as np\r\nimport matplotlib as mpl\r\nimport matplotlib.pyplot as plt\r\n\r\nimg = np.zeros((100, 400))\r\nimg[::5, ::5] = 1\r\nidx = np.indices(img.shape)\r\nidx = [idx[0][::5].flatten(), idx[1][:, ::5].flatten()]\r\n\r\nfig, ax = plt.subplots(constrained_layout=True)\r\nax.imshow(img, interpolation=\"None\", origin=\"lower\")\r\nax.scatter(idx[1], idx[0], marker=\"x\", lw=0.2, s=1)\r\nfig.savefig(\"imshow.pdf\")\n```\n\n\n### Actual outcome\n\nThe image does not align with plot markers, tick marks. Towards the top-right corner, they are even placed in adjacent pixels.\r\n![image](https://user-images.githubusercontent.com/13418822/131690363-80a6fbf1-cd3d-49db-9555-4e49913ff9fb.png)\r\n[Full PDF output](https://github.com/matplotlib/matplotlib/files/7091923/imshow.pdf)\r\n\n\n### Expected outcome\n\nThis can be worked around by calling `_ImageBase._make_image` with `round_to_pixel_border=False`:\r\n\r\n```python\r\nclass NotRoundedAxesImage(mpl.image.AxesImage):\r\n    \"\"\"AxesImage subclass which does not round image extents to fit frame\r\n\r\n    This rounding leads to improper alignment between data and image\r\n    \"\"\"\r\n    def make_image(self, renderer, magnification=1.0, unsampled=False):\r\n        # docstring inherited\r\n        trans = self.get_transform()\r\n        # image is created in the canvas coordinate.\r\n        x1, x2, y1, y2 = self.get_extent()\r\n        bbox = mpl.transforms.Bbox(np.array([[x1, y1], [x2, y2]]))\r\n        transformed_bbox = mpl.transforms.TransformedBbox(bbox, trans)\r\n        clip = ((self.get_clip_box() or self.axes.bbox) if self.get_clip_on()\r\n                else self.figure.bbox)\r\n        # Turn off round_to_pixel_border!\r\n        return self._make_image(self._A, bbox, transformed_bbox, clip,\r\n                                magnification, unsampled=unsampled,\r\n                                round_to_pixel_border=False)\r\n    \r\nfig, ax = plt.subplots(constrained_layout=True)\r\n# Replicate Axes.imshow\r\nax.set_aspect(mpl.rcParams[\"image.aspect\"])\r\nai = NotRoundedAxesImage(ax, interpolation=\"none\")\r\nai.set_data(img)\r\nai.set_extent(ai.get_extent())\r\nax.add_image(ai)\r\nax.scatter(idx[1], idx[0], marker=\"x\", lw=0.2, s=1)\r\nfig.savefig(\"custom_image.pdf\")\r\n```\r\n![image](https://user-images.githubusercontent.com/13418822/131690915-c1eb7166-a75a-4c08-8577-3aa6ad83e4cf.png)\r\n[Full PDF output](https://github.com/matplotlib/matplotlib/files/7091982/custom_image.pdf)\r\n\n\n### Operating system\n\nUbuntu 20.04\n\n### Matplotlib Version\n\n3.4.3\n\n### Matplotlib Backend\n\nPDF, SVG, PGF, probably others\n\n### Python version\n\n3.7\n\n### Jupyter version\n\n_No response_\n\n### Other libraries\n\n_No response_\n\n### Installation\n\nconda\n\n### Conda channel\n\nconda-forge","author":{"url":"https://github.com/lschr","@type":"Person","name":"lschr"},"datePublished":"2021-09-01T14:37:28.000Z","interactionStatistic":{"@type":"InteractionCounter","interactionType":"https://schema.org/CommentAction","userInteractionCount":5},"url":"https://github.com/20964/matplotlib/issues/20964"}