Many hyperlinks are disabled.
Use anonymous login
to enable hyperlinks.
50 most recent check-ins related to "ak-experimental"
|
2011-03-22
| ||
| 17:40 | Made branch "ak-experimental" official, merging it over into the trunk aka main-line. Closing the branch. check-in: cbf1f93ff5 user: andreask tags: trunk | |
|
2011-03-17
| ||
| 21:00 | Simplified package setup and use (demos), by using features from my critcl work (unreleased v2.1). check-in: 13c0452d35 user: andreask tags: work-for-critcl-2.1 | |
| 21:00 | Added buildsystem, brew-style (Derived from critcl). Closed-Leaf check-in: 75bee3e229 user: andreask tags: ak-experimental | |
|
2011-02-20
| ||
| 20:50 | More work on retinexes. Added demo for color retinex working in HSV, processing V like the greyscale retinex, leaving HS unchanged. After a test run on a known set and comparing against the results of GHE the retinex desaturates the colors quite strongly (I suspect that this is the "graying out" noted in the various papers). check-in: e760560b3b user: andreask tags: ak-experimental | |
| 19:58 | Moved some of the supporting code from the retinex demo into the crimp package itself. check-in: bff20fa64c user: andreask tags: ak-experimental | |
| 19:25 | Refactoring the FITFLOAT code a bit to separate the extraction of the necessary statistics a bit. check-in: a1be9797e9 user: andreask tags: ak-experimental | |
| 19:24 | Fixed the unintentional clipping of values found in the float convolutions. Was summing to integer variable, clipping each kernel product, and was also using clamp to grey8 range. Now using a "double" variable for the summing to keep the full precision of all intermediate results, performing any clamping only when storing the final result, and without any clamping at all for images of type "float". The problem was found while experimenting with getting a retinex implementation done, where the middle-range smoothing went spectacularly and visibly wrong because of this. check-in: 43b4914565 user: andreask tags: ak-experimental | |
| 19:24 | Extended the retinex experiment to show the smoothing results. check-in: d5c9bef552 user: andreask tags: ak-experimental | |
|
2011-02-18
| ||
| 06:38 | Experimenting with retinexes, single- and multi-scale, greyscale-only. check-in: ab0ee96f70 user: andreask tags: ak-experimental | |
| 00:16 | Extended the montage operators to handle float images too. check-in: 6d2556f6ac user: andreask tags: ak-experimental | |
|
2011-02-17
| ||
| 20:28 | More fp math ops check-in: d4939e527f user: andreask tags: ak-experimental | |
|
2011-02-09
| ||
| 21:28 | Two bug fixes in the bilaterals: Added the forgotten the scaling factor in the gaussian filtering. And added the missing padding in phase 4, equivalent to the one in phase 2. check-in: d3a262e09b user: andreask tags: ak-experimental | |
| 21:26 | Extended the bilateral demo to show the residuals. check-in: 109ba417bf user: andreask tags: ak-experimental | |
|
2011-01-20
| ||
| 23:54 | Bilateral filter, fixed the striping artifacts. check-in: a2792b8d43 user: andreask tags: ak-experimental | |
| 23:40 | Fix compile, link, and runtime issues with bilateral filter. The filter is definitely incorrect still, has quite obvious defects in the output. check-in: 45378d3264 user: andreask tags: ak-experimental | |
| 23:38 | Added demo for bilateral filter check-in: dd67bbb456 user: andreask tags: ak-experimental | |
|
2011-01-19
| ||
| 23:46 | Joint bilateral filter. Implemented per http://people.csail.mit.edu/sparis/bf_course/. No test cases yet. check-in: 621b077b0f user: andreask tags: ak-experimental | |
| 23:15 | Bilateral filter. Implemented per http://people.csail.mit.edu/sparis/bf_course/. No test cases yet. First use of volume structures. check-in: d03c942360 user: andreask tags: ak-experimental | |
| 23:13 | Integrate volume declarations. check-in: 2c3a7d05fd user: andreask tags: ak-experimental | |
| 23:10 | Added volume structure and basic API. Needed for bilateral grid. Consider re-unification with images, i.e. images == volumes with depth 1. Note also that at least one image processing toolkit uses 4 dimensional structures. x, y, z, and t. check-in: 06c007135c user: andreask tags: ak-experimental | |
|
2011-01-12
| ||
| 19:33 | Updated the List & Has helper commands, now that the C layer forces compilation and loads its shared library there is no need to check the auto_index any longer when looking for primitives. In the future we may also incrementally inline these procedures at their call-sites. check-in: 731ef4e129 user: andreask tags: ak-experimental | |
| 19:30 | Moved the compile check up before we are sourcing the higher API, and modified to actually load the shared library if its ok. The check builds the code, but explicitly isn't loading the results. check-in: 2e71cd9505 user: andreask tags: ak-experimental | |
| 17:50 | Learned how to force the compilation of the C pieces using official/public critcl API calls, and removed the hack from the demos. Plus additional check in crimp if compilation is possible. check-in: b1c7289e2a user: andreask tags: ak-experimental | |
|
2010-12-12
| ||
| 02:27 | Fixed bug in local contrast equalization, added the missing propagation of AHE radius for the rgb/a branches. check-in: 1d1f47843f user: andreask tags: ak-experimental | |
|
2010-12-10
| ||
| 03:48 | Extended gradient demo with a canny edge-detector. check-in: 17abfd8ad4 user: andreask tags: ak-experimental | |
| 03:48 | Additional primitives, the last two steps of canny edge-detection (non-max suppression, and edge-tracing with hysteresis) check-in: 4d11f5d4c9 user: andreask tags: ak-experimental | |
| 02:06 | Documentation update check-in: 5b2297c975 user: andreask tags: ak-experimental | |
|
2010-12-09
| ||
| 22:28 | New demo for the gradient calculation and visualization. check-in: 77ec37b772 user: andreask tags: ak-experimental | |
| 22:25 | Fix more bugs in scharr and prewitt check-in: 20b7bb8e16 user: andreask tags: ak-experimental | |
| 21:15 | Base documentation for the gradient methods and the underlying edge-detecting filters, see [d5d2f1a6f6]. check-in: 565696fdee user: andreask tags: ak-experimental | |
| 19:55 | Base documentation for hypot and atan2 operators. check-in: 050c81b4f1 user: andreask tags: ak-experimental | |
| 08:44 |
Fix bugs in the signature and implementation of the sobel, prewitt, scharr y filters.
Modified the symmetric binary operators to not try argument reversal when the argument image types are identical, as then the second try is identical to the first, thus superfluous. Filled the conversion of gradients from cartesian to polar representation, and added a method to generate a visualization of the gradient. check-in: 1d969fb0c3 user: andreask tags: ak-experimental | |
| 08:01 | Fix oops in name of the primitive. check-in: 48a8abd916 user: andreask tags: ak-experimental | |
| 07:26 | API for the hypot and atan2 operators, see [4dadc8d0a7] and [b2cab740d3] for the underlying primitives. check-in: 859a3e3d61 user: andreask tags: ak-experimental | |
| 07:25 | New binary operator, atan2. Result is always float, regardless of input types check-in: b2cab740d3 user: andreask tags: ak-experimental | |
| 07:02 | Fixed bugs introduced by [a8b237cf02] check-in: f4735e8af6 user: andreask tags: ak-experimental | |
| 07:01 | New binary operator, hypot. Result is always float, regardless of input types check-in: 4dadc8d0a7 user: andreask tags: ak-experimental | |
| 06:59 | Replace custom clamping code with common macro. check-in: 7613b6db74 user: andreask tags: ak-experimental | |
|
2010-12-08
| ||
| 22:09 | Added more predefined filters: sobel, prewitt, scharr. And gradient methods on top of them. check-in: d5d2f1a6f6 user: andreask tags: ak-experimental | |
| 06:40 | Extended the slide show code in demo framework to allow start without cycling the images. check-in: 0e176b570d user: andreask tags: ak-experimental | |
| 06:17 | Added methods computing the kernel of a gaussian blur, discrete, or sampled, and filters based on these kernels. Updated demos and docs. check-in: a8b237cf02 user: andreask tags: ak-experimental | |
| 05:33 | Finally bit the apple and remodified the API of the table eval/linear methods, replacing their option '-wrap' with a deeper submethod, i.e wrap/clamp. The option processing needed to handle -wrap simply forces use of -- always, because of the need to protect negative numbers, and it simply doesn't look nice. Updated implementation, demos, and docs. check-in: 11bc8533d2 user: andreask tags: ak-experimental | |
|
2010-12-07
| ||
| 20:02 | Added more convenience methods for affine transforms: shear, and reflection along lines through the origin. check-in: 9ba5b9ff90 user: andreask tags: ak-experimental | |
| 08:08 | Changed convolution on float image to not clamp the result to grey8 range. Bad. Float is present to allow us the use of large dynamic ranges. check-in: 3d1a041674 user: andreask tags: ak-experimental | |
| 08:07 | Added two additional float operators, hypot and offset (+constant) check-in: 3f72899637 user: andreask tags: ak-experimental | |
| 08:06 | Fix comment of scale float operator, and go double for argument check-in: 3a5d4ff4ec user: andreask tags: ak-experimental | |
| 08:06 | Fix comment of sqrt float operator check-in: c8fc173ab2 user: andreask tags: ak-experimental | |
| 06:47 | Fix bug in expand/const, handle all single-channel image-types. check-in: acc3641e66 user: andreask tags: ak-experimental | |
|
2010-12-02
| ||
| 17:21 | Replaced the Tcl implementation of the reader for PPM files with C code, structured like the C-based PGM reader committed with [186b7b5109]. It simply was stuck in my craw to have a slow PPM reader while knowing that it can be fast too, as demonstrated by the PGM rewrite, and pretty trivial to implement to boot. check-in: 7952dbe0b6 user: andreask tags: ak-experimental | |
|
2010-11-29
| ||
| 01:22 | Fixed bug in [186b7b5109], corner case in handling the end of the PGM header. check-in: c7a7753351 user: andreask tags: ak-experimental | |