r14

• Properly clamp float mask to 0-1 range in mode=1.

r13

• All planes of floating point mask are always in the 0-1 range now.
• Remove TCannyCL filter.
• Add AVX512 and remove AVX.
• Migrate to VS API4.
• Add Kroon, Kirsch and FDoG operators.
• Rename gmmax parameter to scale and change its default to 1.0.

r12

• Add parameter sigma_v.

r11

• TCannyCL: Fix build program failure in some locales.
• TCannyCL: Store the compiled binary for reuse in the offline cache located in $HOME/.boost_compute on UNIX-like systems and in %APPDATA%/boost_compute on Windows.
• TCannyCL: Change filter mode to completely parallel execution.
• Replace Autotools with Meson build system.

r10

• Now the very border pixels are processed rather than zeroed out.
• Remove mode 2 and 3.
• Remove clamping for float sample type in mode -1 and 1.
• Add TCannyCL filter.

r9

• Allow bypassing gaussian blur by setting sigma to 0.

r8

• Change sigma parameter to array type.

r7

• Add opt parameter.

r6

• Add Scharr operator.
• Remove nms parameter.
• Optimize gaussian blur function.
• Vectorize non-maximum suppression function.
• Vectorize output functions of mode -1, 0 and 1.
• Now uses per thread buffer allocation instead of per frame.

r5

• Fix an error in SIMD code of the Sobel operator.