René's URL Explorer Experiment

{"@context":"https://schema.org","@type":"DiscussionForumPosting","headline":"Disappointing Converter behavior","articleBody":"I am starting with this 16-bit image:\r\n\u003cimg width=\"685\" alt=\"image\" src=\"https://github.com/scijava/scijava/assets/1123809/129555b1-8250-4a73-b17e-1ffa587e38e3\"\u003e\r\n\r\nIf I convert it with `Image \u003e Type \u003e 8-bit` in Fiji I get:\r\n\u003cimg width=\"682\" alt=\"image\" src=\"https://github.com/scijava/scijava/assets/1123809/236dfedd-b6a4-4a55-ae9a-e5790d44834d\"\u003e\r\n\r\nIf I denoise this image I get this result (which I'm happy with):\r\n\u003cimg width=\"679\" alt=\"image\" src=\"https://github.com/scijava/scijava/assets/1123809/0adc3d19-7c01-4ca7-b210-399c32dbf98f\"\u003e\r\n\r\nBut if I try to convert it with e.g. `result = ops.unary(\"convert.int8\").input(img).apply()` I get.. a `32-bit` image??\r\n\u003cimg width=\"687\" alt=\"image\" src=\"https://github.com/scijava/scijava/assets/1123809/eb10677c-efb4-4c47-819e-70dc0ef0843f\"\u003e\r\n\r\n`convert.uint8` is 8-bit, at least..\r\n\u003cimg width=\"681\" alt=\"image\" src=\"https://github.com/scijava/scijava/assets/1123809/8597583e-2962-451c-b7d3-b2b4abd5e788\"\u003e\r\n\r\n\r\nSmoothing the uint-8 version requires bumping up the smoothing strength significantly and it still doesn't look great:\r\n\u003cimg width=\"691\" alt=\"image\" src=\"https://github.com/scijava/scijava/assets/1123809/3e13ae29-e027-4dbb-8b22-6457236e080b\"\u003e\r\n\r\nSo, how I account for this behavior is:\r\n- In the ImageJ type conversion, values are scaled proportionately to the new range\r\n- In the SJ Ops type conversion, values are directly translated. This can result in type overflow (note the split tails in both the `int8` and `uint8` conversions) which **fundamentally changes the behavior of subsequent image processing**.\r\n- These [byte casts](https://github.com/scijava/incubator/blob/a969114d2f047a47dffb3ca2d189a3263d56d98a/scijava-ops-image/src/main/java/org/scijava/ops/image/convert/ConvertTypes.java#L111-L125) are perhaps not working as intended?\r\n\r\n@gselzer @ctrueden Is there rationale for not scaling the converters to the target data range?","author":{"url":"https://github.com/hinerm","@type":"Person","name":"hinerm"},"datePublished":"2024-03-12T19:01:50.000Z","interactionStatistic":{"@type":"InteractionCounter","interactionType":"https://schema.org/CommentAction","userInteractionCount":6},"url":"https://github.com/197/scijava/issues/197"}