|
89b14ca9
|
2025-11-25T11:33:25
|
|
change encoding to utf8-unix
|
|
1c8433cb
|
2025-04-10T13:54:58
|
|
Version 3.1: Implemented full C++ compliance, eliminated `stdlib.h` dependency.
Removed a previously added scanline unbiasing since its positive effect was a compiler-dependent statistical error.
Fixed a bug with bit-depth conversions (e.g., 16-bit to 8-bit).
Fixed `sign conversion`, `nullptr`, and other compiler warnings.
Optimized dithering function.
Added 8-bit input gamma correction optimization.
Added optional gamma correction bypass for alpha-channel (`AlphaIndex`).
LANCIR: implemented WebAssembly SIMD128 support.
Improved documentation.
|
|
|
d4884eeb
|
2025-03-31T16:12:32
|
|
LANCIR 3.0.18. Added pre-C++11 compliance.
|
|
|
623bbbc4
|
2025-03-31T15:40:47
|
|
LANCIR 3.0.17. Added pre-C++11 compliance.
Fixed pedantic `sign conversion` and `nullptr` warnings.
|
|
|
84b38194
|
2025-03-24T17:55:39
|
|
LANCIR 3.0.16. Implemented full C++ compliance.
|
|
|
8be60ed2
|
2025-03-24T16:59:48
|
|
LANCIR 3.0.15. Implemented full C++ compliance.
|
|
|
0269b5db
|
2025-03-19T19:28:08
|
|
LANCIR 3.0.14. Improved C++ compliance, comments.
Implemented compile-time branching elimination via template parameters in the outputScanline() function.
Fixed compilation error on ARM32 targets.
Removed redundant ifdef's.
|
|
|
cef36b42
|
2025-03-16T09:35:34
|
|
LANCIR 3.0.13: Added platform detection for MSVC ARM.
|
|
|
eb89c3ba
|
2025-03-15T10:22:13
|
|
LANCIR 3.0.12: Improved platform detection.
|
|
ae50ac8c
|
2024-02-02T17:35:30
|
|
LANCIR 3.0.11: Minor code clean-up.
|
|
|
643fd99e
|
2024-01-18T13:51:54
|
|
LANCIR 3.0.10: Considerably improved large image resizing speed.
Memory overhead now became negligible.
Source and destination buffers should now differ.
Introduced a simpler overloaded resizeImage() function that returns a result.
|
|
|
e928341c
|
2023-12-07T05:40:20
|
|
LANCIR 3.0.9. Changed floating-point `0x1pN` constants not supported by some compilers, to `e` notation.
|
|
|
bb3482a7
|
2023-08-05T10:23:43
|
|
LANCIR 3.0.8. Removed _mm_cvtss_f32 SSSE3 intrinsic dependency.
|
|
|
7271d321
|
2021-12-30T07:22:40
|
|
LANCIR 3.0.7.
Optimized 1-channel resize function for a slightly lower overhead.
|
|
|
1b311293
|
2021-12-28T10:48:48
|
|
LANCIR 3.0.6.
Fixed a bug in update() that returned "true" even if filters were not updated, previously this impacted overhead when resizing many small images.
|
|
|
6e0a1229
|
2021-12-27T20:19:28
|
|
LANCIR 3.0.5.
The "virtual oversampling" has been removed from this version, for an even lower overhead. While it was a correct thing overall, it does not make much of a difference for Lanczos filter. Filter oversampling will likely be used in a future AVIR refactoring to create precise filters while not inducing a computational overhead - it looks like there is no practical need to perform full-featured input image upsizing/oversampling as half of image's samples are initially zero anyway. But for precise filter design oversampling is a necessity, like described on the project page; it has to be tested and analyzed if downsampling a designed oversampled filter works well (it should).
So, LANCIR is now likely has a "production" quality level, its further development is not planned at the moment.
|
|
|
c457bd53
|
2021-12-27T14:45:08
|
|
LANCIR 3.0.4.
Rearranged final summation instructions in `resize3` yielding 2-3% performance boost in 3-channel resizing.
|
|
|
7b89d36d
|
2021-12-26T08:14:13
|
|
LANCIR 3.0.3.
Implemented the "virtual oversampling", at 4X. It is the same technique as originally used in AVIR, but completely overhead-less; improves Lanczos's filter to the maximum.
Optimized filter buffer's allocation.
LANCIR's cutting-edge techniques will be used in a future AVIR refactoring.
|
|
|
3bb2bd95
|
2021-12-24T15:34:07
|
|
LANCIR 3.0.2.
Further algorithm's refinement, now using progressive batched resizing which yields an enormous performance boost in most cases on larger images.
This is a preparation for a major AVIR refactoring.
|
|
|
5809d5b5
|
2021-12-22T15:41:29
|
|
LANCIR 3.0.1. A deep algorithm's redesign. Now features AVX, SSE2, and NEON optimizations out-of-the box.
Now it is likely one of the fastest (if not the fastest) Lanczos algorithms available for CPUs.
NOTE: resizeImage() function's arguments were changed: now also accepting NewBuf's scanline size.
|
|
|
24034240
|
2021-10-03T09:31:06
|
|
Version 3.0.
Improved speed by 10-25% on upsizing by utilizing a special resizing function together with filter-less 2X upsizing. Does not apply to the de-interleaved (AVX) resizing.
Minor LANCIR optimization.
|
|
|
c6e4dead
|
2021-04-15T17:27:09
|
|
Version 2.8.
Fixed regression with the copy-constructor of CImageResizeVars class (previously it caused uninitialized accesses).
Removed filter length optimization as it did not reduce overhead measurably.
Optimized "peaked cosine" window function generator (removed division).
Added "unbiasing" to resizer - an unconventional approach which reduces peak error significantly, at the expense of 5% increased overhead.
Reoptimized filter parameters, now yielding an unprecedented quality.
|
|
|
4efa5c61
|
2021-04-10T22:34:30
|
|
Version 2.7.
Added normalization of individual fractional delay filters. This reduced peak error by 3 dB, which is substantial for image resizing.
Reoptimized all filter parameters resulting in better frequency response linearity.
Added AVIR_NOCTOR macro to avoid copy-constructing and copying objects of some classes via a default copy function.
Added copy-constructor and assignment operator to the CImageResizerVars class, to avoid uninitialized memory copying.
Corrected automatic image offseting and "k" factor on image upsizing.
|
|
|
4ee357d2
|
2021-02-18T15:44:12
|
|
LANCIR: added a missing `ElCount` multiplier to `memset` on empty input image.
|
|
|
fa37d98a
|
2020-03-14T08:45:06
|
|
LANCIR: fixed a rare access violation crash.
|
|
|
96d6501c
|
2019-06-23T22:30:58
|
|
Fixed a crash in CLancIR.
|
|
|
ace3afd9
|
2019-05-31T16:12:09
|
|
Added independent "kx" and "ky" step parameters to the CLancIR resizing function.
|
|
|
f3d8c344
|
2019-05-18T08:42:00
|
|
Made minor text description updates in "lancir".
|
|
|
10e74833
|
2019-05-17T22:44:26
|
|
Added "lancir" filter normalization.
|
|
|
30447d6c
|
2019-05-17T10:03:37
|
|
Made minor optimizations to "lancir".
|
|
|
5ab3db5a
|
2019-05-16T13:01:38
|
|
Made minor "lancir" optimizations.
|
|
|
13d00bb2
|
2019-05-16T06:30:49
|
|
Fixed a minor image offset on image upsizing.
|
|
|
244414db
|
2019-05-16T05:17:44
|
|
Implemented fast CLancIR image resizing algorithm.
|