Many hyperlinks are disabled.
Use anonymous login
to enable hyperlinks.
62 check-ins using file operator/warp-fpcomplex-projective-bilinear.crimp version b64156b415
|
2016-10-27
| ||
| 05:10 | Sync'd with trunk work check-in: 36c7a50438 user: aku tags: refactor-interpolations | |
| 05:08 | Sync'd with trunk work check-in: bc4ac3e851 user: aku tags: double-type | |
| 05:06 | Sync'd with trunk work check-in: e91703a8d4 user: aku tags: testsuite-work | |
|
2014-12-22
| ||
| 19:42 | Created proper image comparison for tests and dropped the -0 it obsoletes from in the matrix inversions tests. Note: No other tests had such. Even so use of the new comparison in more tests may be useful to get away from string equality to proper logical equality. This can be done incrementally. check-in: 7564ebb89f user: andreask tags: refactor-interpolations | |
|
2014-12-20
| ||
| 16:00 | Fix file-name collision for README, on case-insensitive filesystems (Windows, Mac). check-in: f2ee1c72c4 user: andreask tags: refactor-interpolations | |
|
2014-12-11
| ||
| 23:54 | Added tests for (bi)cubic. Changed bicubic implementation to the conceptual form based on cubic as expanded formulas compute something different, sometimes extremely off. check-in: a03f624d1f user: andreask tags: refactor-interpolations | |
| 23:08 | Added testsuite for trilinear interpolation. Passing. check-in: d97d2f4ec7 user: andreask tags: refactor-interpolations | |
| 22:58 | Added testsuites for linear and bilinear interpolation. Passing. check-in: 6798d33c38 user: andreask tags: refactor-interpolations | |
| 22:14 | Added (bi)cubic interpolators. Fixed various issues found by compiler. Added the commands exposig the functionality to script level. Added documentation. Still to do: a testsuite. check-in: 64f8d2a45d user: andreask tags: refactor-interpolations | |
| 01:23 | Put basic interpolation functionality into the core for (re)use by operators. TODO: bicubic implementation, exposure to scripts, documentation, and tests. check-in: 67271fbc1d user: andreask tags: refactor-interpolations | |
| 00:29 | Modified min/max tests to use a strimj-related form for result-specification. check-in: ca452f118e user: andreask tags: double-type | |
| 00:01 | Added tests for min-operator. Tweaked test results for hopefully higher readability. check-in: fcb5538ded user: andreask tags: double-type | |
|
2014-12-10
| ||
| 23:06 | Extended testing of max-operator. check-in: 026bd587d5 user: andreask tags: double-type | |
| 23:05 | Added some missing max/min operators (grey16/32) check-in: a7f2e90741 user: andreask tags: double-type | |
| 21:27 | Reduce precision for trapezoid quadrilateral a bit, found a mismatch at 10. check-in: 7c116931f1 user: andreask tags: double-type | |
| 20:32 | Forced pass of the projective warp tests, by accepting the current results. Not fully happy with that. Thinking about APIs for testing more smaller pieces, i.e. the bilinear and bicubic interpolations, and visibly tracing warp operation. check-in: 017da916e2 user: andreask tags: double-type | |
| 02:03 | Extended the warp-box tests to show the transformed geo corners for more manual validation. check-in: c85c2b0738 user: andreask tags: double-type | |
| 00:59 | Fixed name and descriptions of the failing tests. Still to determine if expected or actual results are wrong. check-in: 18f57a1975 user: andreask tags: double-type | |
| 00:59 | Reworked image display to be have the pixel data vertically aligned. check-in: 4b314e0cd5 user: andreask tags: double-type | |
| 00:16 | Added more core tests for "double"-type. Added accessor "rplace" for doing "relative placement". check-in: f1761672c0 user: andreask tags: double-type | |
|
2014-12-09
| ||
| 23:03 | Fixed warp-box result for quadrilateral, accepting result, given ptransform successes. check-in: d43fa97ece user: andreask tags: double-type | |
| 22:52 | Fixed testing of rotations. Origin and arbitrary rotations now. New polar/cartesian helpers for conversion. Basic checks added before doing the fuzz. check-in: c5960fad26 user: andreask tags: double-type | |
| 00:01 | Extended testsuite to check matrix multiply/inversion for double check-in: 5f0e250a45 user: andreask tags: double-type | |
|
2014-12-08
| ||
| 22:26 | Overhaul of quad testng mostly complete, missing two configs. All pass, fixed bug in p+x command. Rotation failing with changes of previous commit. check-in: 0612fc36dd user: andreask tags: double-type | |
| 21:06 | Tweaks to some projection matrices andd related comments. Fixes of some typos in comments. check-in: 107dc0e3f9 user: aku tags: double-type | |
|
2014-12-06
| ||
| 01:06 | Trialing various quads, currently only the trapezoid code trips the fuzzer, although use of a specific fixed trapezoid did not trip anything. check-in: 544d3794e7 user: andreask tags: double-type | |
| 00:28 | Updated the basic transform fuzz tests to make use of the higher precision afforded by double. Rotations and quadrilaterals still have issues and need (re-)investigation. Quads may simply require generation of proper _convex_ boxes. check-in: 8297dad910 user: andreask tags: double-type | |
|
2014-12-05
| ||
| 20:20 | Update the warp operators to transforms now being of type "double". Tests need updates, results are changed with the higher precision. check-in: ee64da3903 user: andreask tags: double-type | |
| 20:00 | Fix oops in write::tcl::double, still had a float in there. check-in: eaa07448f7 user: andreask tags: double-type | |
| 19:58 | Tests, buglet fix in bwv, plus update of the transform checker. check-in: 400fe45089 user: andreask tags: double-type | |
| 19:51 | Moved transformation matrices for warp over to type "double" for proper precision. check-in: 9948dee714 user: andreask tags: double-type | |
| 19:49 | Added the missing linearmap functions involving double check-in: d048d2b809 user: andreask tags: double-type | |
| 09:14 | Fix build errors. check-in: 2518736e62 user: aku tags: double-type | |
| 09:02 | Fix a few comment typos check-in: c9280dbc25 user: aku tags: double-type | |
| 09:01 | Expanded the set of operations accepting/returning "double"-images. check-in: 1ef28b7f6a user: aku tags: double-type | |
| 02:18 | Started double type support, for higher precision warping. Projection matrices shall be double, always. check-in: 9e3c292ca9 user: andreask tags: double-type | |
|
2013-02-01
| ||
| 08:26 | Documentation: Updated location of critcl, regenerated embedded. check-in: 190bc2a654 user: andreask tags: testsuite-work | |
|
2012-09-14
| ||
| 03:02 | Fixed bug in revision [f255f089dfbb75cbdfeb], dropped the origin assignments in the refactoring. Any transform used random values from uninitialized memory. check-in: 5403a91b50 user: andreask tags: testsuite-work | |
| 02:31 | Tweaked test recipe, capture stderr. check-in: f640f9dcfe user: andreask tags: testsuite-work | |
|
2012-06-29
| ||
| 23:19 | Tweaked test logging check-in: 95ad6ba23a user: andreask tags: testsuite-work | |
| 23:19 | Fixed bug in geometry warping, off-by-one check-in: eac8738308 user: andreask tags: testsuite-work | |
| 23:18 | Added tests for geometry warping, variable cleanup through apply scopes. check-in: 92d0204c43 user: andreask tags: testsuite-work | |
| 20:50 | Moved warp-base testing to custom matching and added the quadrilateral. check-in: d652011fd7 user: andreask tags: testsuite-work | |
| 19:03 | Purged definition and use of [origin] from the testsuite check-in: 572bed05ac user: andreask tags: testsuite-work | |
| 18:54 | Put some of my external helper code into the 'test' command of build.tcl (save full log, show fails directly, collect and show statistics at the end). Disabled the 'origin' helper command, replaced by use of '-verbose line' we have in tcltest already (2.3+). check-in: 44141b1673 user: andreask tags: testsuite-work | |
| 00:06 | Fixed stupid bug in reflection about the axes. Switched all transforms tests over to decode_transform + custom match. check-in: cb7d8233da user: andreask tags: testsuite-work | |
|
2012-06-28
| ||
| 23:42 | Seem to have figured out the quads. The paper I was working from seems to have one bug, and one notational issue which wasn't clear. quads emulating most regular transforms now generate the proper matrices of these transforms. Fuzz testing still disabled, still problematic. Likely because the transform assumes convex quads and the random boxes I generate pretty much are not. check-in: ceceb5890f user: andreask tags: testsuite-work | |
| 22:00 | Continued chipping at the quad transform. Worked out a better way for matching transform results (decode + numeric equality, much simpler result spec). Fixed an apparent ordering issue in quad transform result. Added randomized tests using the Q2UNIT core. Disabled at the moment. Transform is still buggy (identity ok, translation already not). check-in: 65f036813d user: andreask tags: testsuite-work | |
| 18:00 | Extended scaling to allow for arbitrary center of scaling (like for rotation). Updated testsuite. check-in: be0529d8de user: andreask tags: testsuite-work | |
| 17:15 | Fixed implementation of shear transform. Annotated everything with the respective matrices as proper reminders. Updated testsuite a bit. Random testing of shear now passes. check-in: ef4d69a97e user: andreask tags: testsuite-work | |
| 00:06 | Added randomized testing for shear. This is known buggy in either transform, or the test shear setup. check-in: 3896c3d5ae user: andreask tags: testsuite-work | |
|
2012-06-27
| ||
| 23:02 | Added randomized testing for rotation. This seems to be getting hit hard by single-precision math. Manage to match only within 1e1 (before! decimal point). Really consider some way of implementing double-precision math for transforms (at least ?!). check-in: 1a3ee90bf4 user: andreask tags: testsuite-work | |
| 22:43 | Adding randomized testing of translation, scaling and arbitrary reflection. check-in: 1f6667ec08 user: andreask tags: testsuite-work | |
| 20:43 | Testsuite helpers extended with numeric match commands, random number/point generation, and 2d vector arithmetics. check-in: d09baea48e user: andreask tags: testsuite-work | |
|
2012-06-26
| ||
| 05:51 | More annotations for the low-level matrix and vector arithmetics. Simplified the chain implementation, plus annotations of its workings. Updated documentation. Regenerated embedded docs. check-in: d4db707423 user: andreask tags: testsuite-work | |
| 00:43 | Straigthened out transformation application order issues found during manual verification of transform results. check-in: 8d1850a713 user: andreask tags: testsuite-work | |
|
2012-06-25
| ||
| 23:44 | Extended testsuite for 'warp point', scaling, translation. Ok. check-in: 54db908601 user: andreask tags: testsuite-work | |
| 23:40 | Extended testsuite for 'warp point', reflection, rotation. Needs manual verification of results. check-in: ce65c1fa86 user: andreask tags: testsuite-work | |
| 23:19 | Nicer argument names for the new warp methods. Testsuite extended, basic argument checking of the new methods. check-in: 6a68425aff user: andreask tags: testsuite-work | |
| 23:05 | Refactored the geometry setup for warping a bit, and exposed as Tcl commands for use by the testsuite. check-in: f255f089df user: andreask tags: testsuite-work | |
|
2012-06-23
| ||
| 00:07 | Started on test for 'warp projective'. Not looking nice. Have to work on and test the individual pieces of the whole pipeline, like transforming the image corners, backward transform, generate a warp field from a projective transform. check-in: 4e9f664984 user: andreask tags: testsuite-work | |
|
2012-06-22
| ||
| 23:41 | Extended 'warp projective' to handle image type fpcomplex. check-in: bbb1d2c911 user: andreask tags: testsuite-work | |