imgproc_gpu.cpp 35.5 KB
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 61 62 63 64 65 66 67 68 69 70 71 72 73 74 75 76 77 78 79 80 81 82 83 84 85 86 87 88 89 90 91 92 93 94 95 96 97 98 99 100 101 102 103 104 105 106 107 108 109 110 111 112 113 114 115 116 117 118 119 120 121 122 123 124 125 126 127 128 129 130 131 132 133 134 135 136 137 138 139 140 141 142 143 144 145 146 147 148 149 150 151 152 153 154 155 156 157 158 159 160 161 162 163 164 165 166 167 168 169 170 171 172 173 174 175 176 177 178 179 180 181 182 183 184 185 186 187 188 189 190 191 192 193 194 195 196 197 198 199 200 201 202 203 204 205 206 207 208 209 210 211 212 213 214 215 216 217 218 219 220 221 222 223 224 225 226 227 228 229 230 231 232 233 234 235 236 237 238 239 240 241 242 243 244 245 246 247 248 249 250 251 252 253 254 255 256 257 258 259 260 261 262 263 264 265 266 267 268 269 270 271 272 273 274 275 276 277 278 279 280 281 282 283 284 285 286 287 288 289 290 291 292 293 294 295 296 297 298 299 300 301 302 303 304 305 306 307 308 309 310 311 312 313 314 315 316 317 318 319 320 321 322 323 324 325 326 327 328 329 330 331 332 333 334 335 336 337 338 339 340 341 342 343 344 345 346 347 348 349 350 351 352 353 354 355 356 357 358 359 360 361 362 363 364 365 366 367 368 369 370 371 372 373 374 375 376 377 378 379 380 381 382 383 384 385 386 387 388 389 390 391 392 393 394 395 396 397 398 399 400 401 402 403 404 405 406 407 408 409 410 411 412 413 414 415 416 417 418 419 420 421 422 423 424 425 426 427 428 429 430 431 432 433 434 435 436 437 438 439 440 441 442 443 444 445 446 447 448 449 450 451 452 453 454 455 456 457 458 459 460 461 462 463 464 465 466 467 468 469 470 471 472 473 474 475 476 477 478 479 480 481 482 483 484 485 486 487 488 489 490 491 492 493 494 495 496 497 498 499 500 501 502 503 504 505 506 507 508 509 510 511 512 513 514 515 516 517 518 519 520 521 522 523 524 525 526 527 528 529 530 531 532 533 534 535 536 537 538 539 540 541 542 543 544 545 546 547 548 549 550 551 552 553 554 555 556 557 558 559 560 561 562 563 564 565 566 567 568 569 570 571 572 573 574 575 576 577 578 579 580 581 582 583 584 585 586 587 588 589 590 591 592 593 594 595 596 597 598 599 600 601 602 603 604 605 606 607 608 609 610 611 612 613 614 615 616 617 618 619 620 621 622 623 624 625 626 627 628 629 630 631 632 633 634 635 636 637 638 639 640 641 642 643 644 645 646 647 648 649 650 651 652 653 654 655 656 657 658 659 660 661 662 663 664 665 666 667 668 669 670 671 672 673 674 675 676 677 678 679 680 681 682 683 684 685 686 687 688 689 690 691 692 693 694 695 696 697 698 699 700 701 702 703 704 705 706 707 708 709 710 711 712 713 714 715 716 717 718 719 720 721 722 723 724 725 726 727 728 729 730 731 732 733 734 735 736 737 738 739 740 741 742 743 744 745 746 747 748 749 750 751 752 753 754 755 756 757 758 759 760 761 762 763 764 765 766 767 768 769 770 771 772 773 774 775 776 777 778 779 780 781 782 783 784 785 786 787 788 789 790 791 792 793 794 795 796 797 798 799 800 801 802 803 804 805 806 807 808 809 810 811 812 813 814 815 816 817 818 819 820 821 822 823 824 825 826 827 828 829 830 831 832 833 834 835 836 837 838 839 840 841 842 843 844 845 846 847 848 849 850 851 852 853 854 855 856 857 858 859 860 861 862 863 864 865 866 867 868 869 870 871 872 873 874 875 876 877 878 879 880 881 882 883 884 885
/*M///////////////////////////////////////////////////////////////////////////////////////
//
//  IMPORTANT: READ BEFORE DOWNLOADING, COPYING, INSTALLING OR USING.
//
//  By downloading, copying, installing or using the software you agree to this license.
//  If you do not agree to this license, do not download, install,
//  copy or use the software.
//
//
//                           License Agreement
//                For Open Source Computer Vision Library
//
// Copyright (C) 2000-2008, Intel Corporation, all rights reserved.
// Copyright (C) 2009, Willow Garage Inc., all rights reserved.
// Third party copyrights are property of their respective owners.
//
// Redistribution and use in source and binary forms, with or without modification,
// are permitted provided that the following conditions are met:
//
//   * Redistribution's of source code must retain the above copyright notice,
//     this list of conditions and the following disclaimer.
//
//   * Redistribution's in binary form must reproduce the above copyright notice,
//     this list of conditions and the following disclaimer in the documentation
//     and/or other materials provided with the distribution.
//
//   * The name of the copyright holders may not be used to endorse or promote products
//     derived from this software without specific prior written permission.
//
// This software is provided by the copyright holders and contributors "as is" and
// any express or implied warranties, including, but not limited to, the implied
// warranties of merchantability and fitness for a particular purpose are disclaimed.
// In no event shall the Intel Corporation or contributors be liable for any direct,
// indirect, incidental, special, exemplary, or consequential damages
// (including, but not limited to, procurement of substitute goods or services;
// loss of use, data, or profits; or business interruption) however caused
// and on any theory of liability, whether in contract, strict liability,
// or tort (including negligence or otherwise) arising in any way out of
// the use of this software, even if advised of the possibility of such damage.
//
//M*/

#include "precomp.hpp"

using namespace cv;
using namespace cv::gpu;

#if !defined (HAVE_CUDA)

void cv::gpu::remap(const GpuMat&, GpuMat&, const GpuMat&, const GpuMat&){ throw_nogpu(); }
void cv::gpu::meanShiftFiltering(const GpuMat&, GpuMat&, int, int, TermCriteria) { throw_nogpu(); }
void cv::gpu::meanShiftProc(const GpuMat&, GpuMat&, GpuMat&, int, int, TermCriteria) { throw_nogpu(); }
void cv::gpu::drawColorDisp(const GpuMat&, GpuMat&, int) { throw_nogpu(); }
void cv::gpu::drawColorDisp(const GpuMat&, GpuMat&, int, const Stream&) { throw_nogpu(); }
void cv::gpu::reprojectImageTo3D(const GpuMat&, GpuMat&, const Mat&) { throw_nogpu(); }
void cv::gpu::reprojectImageTo3D(const GpuMat&, GpuMat&, const Mat&, const Stream&) { throw_nogpu(); }
double cv::gpu::threshold(const GpuMat&, GpuMat&, double) { throw_nogpu(); return 0.0; }
void cv::gpu::resize(const GpuMat&, GpuMat&, Size, double, double, int) { throw_nogpu(); }
void cv::gpu::copyMakeBorder(const GpuMat&, GpuMat&, int, int, int, int, const Scalar&) { throw_nogpu(); }
void cv::gpu::warpAffine(const GpuMat&, GpuMat&, const Mat&, Size, int) { throw_nogpu(); }
void cv::gpu::warpPerspective(const GpuMat&, GpuMat&, const Mat&, Size, int) { throw_nogpu(); }
void cv::gpu::rotate(const GpuMat&, GpuMat&, Size, double, double, double, int) { throw_nogpu(); }
void cv::gpu::integral(GpuMat&, GpuMat&, GpuMat&) { throw_nogpu(); }
void cv::gpu::rectStdDev(const GpuMat&, const GpuMat&, GpuMat&, const Rect&) { throw_nogpu(); }
void cv::gpu::Canny(const GpuMat&, GpuMat&, double, double, int) { throw_nogpu(); }
void cv::gpu::evenLevels(GpuMat&, int, int, int) { throw_nogpu(); }
void cv::gpu::histEven(const GpuMat&, GpuMat&, int, int, int) { throw_nogpu(); }
void cv::gpu::histEven(const GpuMat&, GpuMat*, int*, int*, int*) { throw_nogpu(); }
void cv::gpu::histRange(const GpuMat&, GpuMat&, const GpuMat&) { throw_nogpu(); }
void cv::gpu::histRange(const GpuMat&, GpuMat*, const GpuMat*) { throw_nogpu(); }

#else /* !defined (HAVE_CUDA) */

#define NPP_VERSION (10 * NPP_VERSION_MAJOR + NPP_VERSION_MINOR)

namespace cv { namespace gpu {  namespace imgproc
{
    void remap_gpu_1c(const DevMem2D& src, const DevMem2Df& xmap, const DevMem2Df& ymap, DevMem2D dst);
    void remap_gpu_3c(const DevMem2D& src, const DevMem2Df& xmap, const DevMem2Df& ymap, DevMem2D dst);

    extern "C" void meanShiftFiltering_gpu(const DevMem2D& src, DevMem2D dst, int sp, int sr, int maxIter, float eps);
    extern "C" void meanShiftProc_gpu(const DevMem2D& src, DevMem2D dstr, DevMem2D dstsp, int sp, int sr, int maxIter, float eps);

    void drawColorDisp_gpu(const DevMem2D& src, const DevMem2D& dst, int ndisp, const cudaStream_t& stream);
    void drawColorDisp_gpu(const DevMem2D_<short>& src, const DevMem2D& dst, int ndisp, const cudaStream_t& stream);

    void reprojectImageTo3D_gpu(const DevMem2D& disp, const DevMem2Df& xyzw, const float* q, const cudaStream_t& stream);
    void reprojectImageTo3D_gpu(const DevMem2D_<short>& disp, const DevMem2Df& xyzw, const float* q, const cudaStream_t& stream);
}}}

////////////////////////////////////////////////////////////////////////
// remap

void cv::gpu::remap(const GpuMat& src, GpuMat& dst, const GpuMat& xmap, const GpuMat& ymap)
{
    typedef void (*remap_gpu_t)(const DevMem2D& src, const DevMem2Df& xmap, const DevMem2Df& ymap, DevMem2D dst);
    static const remap_gpu_t callers[] = {imgproc::remap_gpu_1c, 0, imgproc::remap_gpu_3c};

    CV_Assert((src.type() == CV_8U || src.type() == CV_8UC3) && xmap.type() == CV_32F && ymap.type() == CV_32F);

    GpuMat out;
    if (dst.data != src.data)
        out = dst;

    out.create(xmap.size(), src.type());

    callers[src.channels() - 1](src, xmap, ymap, out);

    dst = out;
}

////////////////////////////////////////////////////////////////////////
// meanShiftFiltering_GPU

void cv::gpu::meanShiftFiltering(const GpuMat& src, GpuMat& dst, int sp, int sr, TermCriteria criteria)
{
    if( src.empty() )
        CV_Error( CV_StsBadArg, "The input image is empty" );

    if( src.depth() != CV_8U || src.channels() != 4 )
        CV_Error( CV_StsUnsupportedFormat, "Only 8-bit, 4-channel images are supported" );

    dst.create( src.size(), CV_8UC4 );

    if( !(criteria.type & TermCriteria::MAX_ITER) )
        criteria.maxCount = 5;

    int maxIter = std::min(std::max(criteria.maxCount, 1), 100);

    float eps;
    if( !(criteria.type & TermCriteria::EPS) )
        eps = 1.f;
    eps = (float)std::max(criteria.epsilon, 0.0);

    imgproc::meanShiftFiltering_gpu(src, dst, sp, sr, maxIter, eps);
}

////////////////////////////////////////////////////////////////////////
// meanShiftProc_GPU

void cv::gpu::meanShiftProc(const GpuMat& src, GpuMat& dstr, GpuMat& dstsp, int sp, int sr, TermCriteria criteria)
{
    if( src.empty() )
        CV_Error( CV_StsBadArg, "The input image is empty" );

    if( src.depth() != CV_8U || src.channels() != 4 )
        CV_Error( CV_StsUnsupportedFormat, "Only 8-bit, 4-channel images are supported" );

    dstr.create( src.size(), CV_8UC4 );
    dstsp.create( src.size(), CV_16SC2 );

    if( !(criteria.type & TermCriteria::MAX_ITER) )
        criteria.maxCount = 5;

    int maxIter = std::min(std::max(criteria.maxCount, 1), 100);

    float eps;
    if( !(criteria.type & TermCriteria::EPS) )
        eps = 1.f;
    eps = (float)std::max(criteria.epsilon, 0.0);

    imgproc::meanShiftProc_gpu(src, dstr, dstsp, sp, sr, maxIter, eps);
}

////////////////////////////////////////////////////////////////////////
// drawColorDisp

namespace
{
    template <typename T>
    void drawColorDisp_caller(const GpuMat& src, GpuMat& dst, int ndisp, const cudaStream_t& stream)
    {
        GpuMat out;
        if (dst.data != src.data)
            out = dst;
        out.create(src.size(), CV_8UC4);

        imgproc::drawColorDisp_gpu((DevMem2D_<T>)src, out, ndisp, stream);

        dst = out;
    }

    typedef void (*drawColorDisp_caller_t)(const GpuMat& src, GpuMat& dst, int ndisp, const cudaStream_t& stream);

    const drawColorDisp_caller_t drawColorDisp_callers[] = {drawColorDisp_caller<unsigned char>, 0, 0, drawColorDisp_caller<short>, 0, 0, 0, 0};
}

void cv::gpu::drawColorDisp(const GpuMat& src, GpuMat& dst, int ndisp)
{
    CV_Assert(src.type() == CV_8U || src.type() == CV_16S);

    drawColorDisp_callers[src.type()](src, dst, ndisp, 0);
}

void cv::gpu::drawColorDisp(const GpuMat& src, GpuMat& dst, int ndisp, const Stream& stream)
{
    CV_Assert(src.type() == CV_8U || src.type() == CV_16S);

    drawColorDisp_callers[src.type()](src, dst, ndisp, StreamAccessor::getStream(stream));
}

////////////////////////////////////////////////////////////////////////
// reprojectImageTo3D

namespace
{
    template <typename T>
    void reprojectImageTo3D_caller(const GpuMat& disp, GpuMat& xyzw, const Mat& Q, const cudaStream_t& stream)
    {
        xyzw.create(disp.rows, disp.cols, CV_32FC4);
        imgproc::reprojectImageTo3D_gpu((DevMem2D_<T>)disp, xyzw, Q.ptr<float>(), stream);
    }

    typedef void (*reprojectImageTo3D_caller_t)(const GpuMat& disp, GpuMat& xyzw, const Mat& Q, const cudaStream_t& stream);

    const reprojectImageTo3D_caller_t reprojectImageTo3D_callers[] = {reprojectImageTo3D_caller<unsigned char>, 0, 0, reprojectImageTo3D_caller<short>, 0, 0, 0, 0};
}

void cv::gpu::reprojectImageTo3D(const GpuMat& disp, GpuMat& xyzw, const Mat& Q)
{
    CV_Assert((disp.type() == CV_8U || disp.type() == CV_16S) && Q.type() == CV_32F && Q.rows == 4 && Q.cols == 4);

    reprojectImageTo3D_callers[disp.type()](disp, xyzw, Q, 0);
}

void cv::gpu::reprojectImageTo3D(const GpuMat& disp, GpuMat& xyzw, const Mat& Q, const Stream& stream)
{
    CV_Assert((disp.type() == CV_8U || disp.type() == CV_16S) && Q.type() == CV_32F && Q.rows == 4 && Q.cols == 4);

    reprojectImageTo3D_callers[disp.type()](disp, xyzw, Q, StreamAccessor::getStream(stream));
}

////////////////////////////////////////////////////////////////////////
// threshold

double cv::gpu::threshold(const GpuMat& src, GpuMat& dst, double thresh)
{
    CV_Assert(src.type() == CV_32FC1);

    dst.create( src.size(), src.type() );

    NppiSize sz;
    sz.width  = src.cols;
    sz.height = src.rows;

    nppSafeCall( nppiThreshold_32f_C1R(src.ptr<Npp32f>(), src.step,
        dst.ptr<Npp32f>(), dst.step, sz, static_cast<Npp32f>(thresh), NPP_CMP_GREATER) );

    return thresh;
}

////////////////////////////////////////////////////////////////////////
// resize

void cv::gpu::resize(const GpuMat& src, GpuMat& dst, Size dsize, double fx, double fy, int interpolation)
{
    static const int npp_inter[] = {NPPI_INTER_NN, NPPI_INTER_LINEAR/*, NPPI_INTER_CUBIC, 0, NPPI_INTER_LANCZOS*/};

    CV_Assert(src.type() == CV_8UC1 || src.type() == CV_8UC4);
    CV_Assert(interpolation == INTER_NEAREST || interpolation == INTER_LINEAR/* || interpolation == INTER_CUBIC || interpolation == INTER_LANCZOS4*/);

    CV_Assert( src.size().area() > 0 );
    CV_Assert( !(dsize == Size()) || (fx > 0 && fy > 0) );

    if( dsize == Size() )
    {
        dsize = Size(saturate_cast<int>(src.cols * fx), saturate_cast<int>(src.rows * fy));
    }
    else
    {
        fx = (double)dsize.width / src.cols;
        fy = (double)dsize.height / src.rows;
    }

    dst.create(dsize, src.type());

    NppiSize srcsz;
    srcsz.width  = src.cols;
    srcsz.height = src.rows;
    NppiRect srcrect;
    srcrect.x = srcrect.y = 0;
    srcrect.width  = src.cols;
    srcrect.height = src.rows;
    NppiSize dstsz;
    dstsz.width  = dst.cols;
    dstsz.height = dst.rows;

    if (src.type() == CV_8UC1)
    {
        nppSafeCall( nppiResize_8u_C1R(src.ptr<Npp8u>(), srcsz, src.step, srcrect,
            dst.ptr<Npp8u>(), dst.step, dstsz, fx, fy, npp_inter[interpolation]) );
    }
    else
    {
        nppSafeCall( nppiResize_8u_C4R(src.ptr<Npp8u>(), srcsz, src.step, srcrect,
            dst.ptr<Npp8u>(), dst.step, dstsz, fx, fy, npp_inter[interpolation]) );
    }
}

////////////////////////////////////////////////////////////////////////
// copyMakeBorder

void cv::gpu::copyMakeBorder(const GpuMat& src, GpuMat& dst, int top, int bottom, int left, int right, const Scalar& value)
{
    CV_Assert(src.type() == CV_8UC1 || src.type() == CV_8UC4 || src.type() == CV_32SC1);

    dst.create(src.rows + top + bottom, src.cols + left + right, src.type());

	NppiSize srcsz;
	srcsz.width  = src.cols;
	srcsz.height = src.rows;
    NppiSize dstsz;
        dstsz.width  = dst.cols;
        dstsz.height = dst.rows;

    switch (src.type())
    {
    case CV_8UC1:
                {
            Npp8u nVal = static_cast<Npp8u>(value[0]);
            nppSafeCall( nppiCopyConstBorder_8u_C1R(src.ptr<Npp8u>(), src.step, srcsz,
                dst.ptr<Npp8u>(), dst.step, dstsz, top, left, nVal) );
            break;
                }
    case CV_8UC4:
                {
            Npp8u nVal[] = {static_cast<Npp8u>(value[0]), static_cast<Npp8u>(value[1]), static_cast<Npp8u>(value[2]), static_cast<Npp8u>(value[3])};
            nppSafeCall( nppiCopyConstBorder_8u_C4R(src.ptr<Npp8u>(), src.step, srcsz,
                dst.ptr<Npp8u>(), dst.step, dstsz, top, left, nVal) );
            break;
                }
    case CV_32SC1:
            {
            Npp32s nVal = static_cast<Npp32s>(value[0]);
            nppSafeCall( nppiCopyConstBorder_32s_C1R(src.ptr<Npp32s>(), src.step, srcsz,
                dst.ptr<Npp32s>(), dst.step, dstsz, top, left, nVal) );
            break;
            }
    default:
        CV_Assert(!"Unsupported source type");
    }
}

////////////////////////////////////////////////////////////////////////
// warp

namespace
{
    typedef NppStatus (*npp_warp_8u_t)(const Npp8u* pSrc, NppiSize srcSize, int srcStep, NppiRect srcRoi, Npp8u* pDst,
                                       int dstStep, NppiRect dstRoi, const double coeffs[][3],
                                       int interpolation);
    typedef NppStatus (*npp_warp_16u_t)(const Npp16u* pSrc, NppiSize srcSize, int srcStep, NppiRect srcRoi, Npp16u* pDst,
                                       int dstStep, NppiRect dstRoi, const double coeffs[][3],
                                       int interpolation);
    typedef NppStatus (*npp_warp_32s_t)(const Npp32s* pSrc, NppiSize srcSize, int srcStep, NppiRect srcRoi, Npp32s* pDst,
                                       int dstStep, NppiRect dstRoi, const double coeffs[][3],
                                       int interpolation);
    typedef NppStatus (*npp_warp_32f_t)(const Npp32f* pSrc, NppiSize srcSize, int srcStep, NppiRect srcRoi, Npp32f* pDst,
                                       int dstStep, NppiRect dstRoi, const double coeffs[][3],
                                       int interpolation);

    void nppWarpCaller(const GpuMat& src, GpuMat& dst, double coeffs[][3], const Size& dsize, int flags,
                       npp_warp_8u_t npp_warp_8u[][2], npp_warp_16u_t npp_warp_16u[][2],
                       npp_warp_32s_t npp_warp_32s[][2], npp_warp_32f_t npp_warp_32f[][2])
    {
        static const int npp_inter[] = {NPPI_INTER_NN, NPPI_INTER_LINEAR, NPPI_INTER_CUBIC};

        int interpolation = flags & INTER_MAX;

        CV_Assert((src.depth() == CV_8U || src.depth() == CV_16U || src.depth() == CV_32S || src.depth() == CV_32F) && src.channels() != 2);
        CV_Assert(interpolation == INTER_NEAREST || interpolation == INTER_LINEAR || interpolation == INTER_CUBIC);

        dst.create(dsize, src.type());

        NppiSize srcsz;
        srcsz.height = src.rows;
        srcsz.width = src.cols;
        NppiRect srcroi;
        srcroi.x = srcroi.y = 0;
        srcroi.height = src.rows;
        srcroi.width = src.cols;
        NppiRect dstroi;
        dstroi.x = dstroi.y = 0;
        dstroi.height = dst.rows;
        dstroi.width = dst.cols;

        int warpInd = (flags & WARP_INVERSE_MAP) >> 4;

        switch (src.depth())
        {
        case CV_8U:
            nppSafeCall( npp_warp_8u[src.channels()][warpInd](src.ptr<Npp8u>(), srcsz, src.step, srcroi,
                dst.ptr<Npp8u>(), dst.step, dstroi, coeffs, npp_inter[interpolation]) );
            break;
        case CV_16U:
            nppSafeCall( npp_warp_16u[src.channels()][warpInd](src.ptr<Npp16u>(), srcsz, src.step, srcroi,
                dst.ptr<Npp16u>(), dst.step, dstroi, coeffs, npp_inter[interpolation]) );
            break;
        case CV_32S:
            nppSafeCall( npp_warp_32s[src.channels()][warpInd](src.ptr<Npp32s>(), srcsz, src.step, srcroi,
                dst.ptr<Npp32s>(), dst.step, dstroi, coeffs, npp_inter[interpolation]) );
            break;
        case CV_32F:
            nppSafeCall( npp_warp_32f[src.channels()][warpInd](src.ptr<Npp32f>(), srcsz, src.step, srcroi,
                dst.ptr<Npp32f>(), dst.step, dstroi, coeffs, npp_inter[interpolation]) );
            break;
        default:
            CV_Assert(!"Unsupported source type");
        }
    }
}

void cv::gpu::warpAffine(const GpuMat& src, GpuMat& dst, const Mat& M, Size dsize, int flags)
{
    static npp_warp_8u_t npp_warpAffine_8u[][2] =
        {
            {0, 0},
            {nppiWarpAffine_8u_C1R, nppiWarpAffineBack_8u_C1R},
            {0, 0},
            {nppiWarpAffine_8u_C3R, nppiWarpAffineBack_8u_C3R},
            {nppiWarpAffine_8u_C4R, nppiWarpAffineBack_8u_C4R}
        };
    static npp_warp_16u_t npp_warpAffine_16u[][2] =
        {
            {0, 0},
            {nppiWarpAffine_16u_C1R, nppiWarpAffineBack_16u_C1R},
            {0, 0},
            {nppiWarpAffine_16u_C3R, nppiWarpAffineBack_16u_C3R},
            {nppiWarpAffine_16u_C4R, nppiWarpAffineBack_16u_C4R}
        };
    static npp_warp_32s_t npp_warpAffine_32s[][2] =
        {
            {0, 0},
            {nppiWarpAffine_32s_C1R, nppiWarpAffineBack_32s_C1R},
            {0, 0},
            {nppiWarpAffine_32s_C3R, nppiWarpAffineBack_32s_C3R},
            {nppiWarpAffine_32s_C4R, nppiWarpAffineBack_32s_C4R}
        };
    static npp_warp_32f_t npp_warpAffine_32f[][2] =
        {
            {0, 0},
            {nppiWarpAffine_32f_C1R, nppiWarpAffineBack_32f_C1R},
            {0, 0},
            {nppiWarpAffine_32f_C3R, nppiWarpAffineBack_32f_C3R},
            {nppiWarpAffine_32f_C4R, nppiWarpAffineBack_32f_C4R}
        };

    CV_Assert(M.rows == 2 && M.cols == 3);

    double coeffs[2][3];
    Mat coeffsMat(2, 3, CV_64F, (void*)coeffs);
    M.convertTo(coeffsMat, coeffsMat.type());

    nppWarpCaller(src, dst, coeffs, dsize, flags, npp_warpAffine_8u, npp_warpAffine_16u, npp_warpAffine_32s, npp_warpAffine_32f);
}

void cv::gpu::warpPerspective(const GpuMat& src, GpuMat& dst, const Mat& M, Size dsize, int flags)
{
    static npp_warp_8u_t npp_warpPerspective_8u[][2] =
        {
            {0, 0},
            {nppiWarpPerspective_8u_C1R, nppiWarpPerspectiveBack_8u_C1R},
            {0, 0},
            {nppiWarpPerspective_8u_C3R, nppiWarpPerspectiveBack_8u_C3R},
            {nppiWarpPerspective_8u_C4R, nppiWarpPerspectiveBack_8u_C4R}
        };
    static npp_warp_16u_t npp_warpPerspective_16u[][2] =
        {
            {0, 0},
            {nppiWarpPerspective_16u_C1R, nppiWarpPerspectiveBack_16u_C1R},
            {0, 0},
            {nppiWarpPerspective_16u_C3R, nppiWarpPerspectiveBack_16u_C3R},
            {nppiWarpPerspective_16u_C4R, nppiWarpPerspectiveBack_16u_C4R}
        };
    static npp_warp_32s_t npp_warpPerspective_32s[][2] =
        {
            {0, 0},
            {nppiWarpPerspective_32s_C1R, nppiWarpPerspectiveBack_32s_C1R},
            {0, 0},
            {nppiWarpPerspective_32s_C3R, nppiWarpPerspectiveBack_32s_C3R},
            {nppiWarpPerspective_32s_C4R, nppiWarpPerspectiveBack_32s_C4R}
        };
    static npp_warp_32f_t npp_warpPerspective_32f[][2] =
        {
            {0, 0},
            {nppiWarpPerspective_32f_C1R, nppiWarpPerspectiveBack_32f_C1R},
            {0, 0},
            {nppiWarpPerspective_32f_C3R, nppiWarpPerspectiveBack_32f_C3R},
            {nppiWarpPerspective_32f_C4R, nppiWarpPerspectiveBack_32f_C4R}
        };

    CV_Assert(M.rows == 3 && M.cols == 3);

    double coeffs[3][3];
    Mat coeffsMat(3, 3, CV_64F, (void*)coeffs);
    M.convertTo(coeffsMat, coeffsMat.type());

    nppWarpCaller(src, dst, coeffs, dsize, flags, npp_warpPerspective_8u, npp_warpPerspective_16u, npp_warpPerspective_32s, npp_warpPerspective_32f);
}

////////////////////////////////////////////////////////////////////////
// rotate

void cv::gpu::rotate(const GpuMat& src, GpuMat& dst, Size dsize, double angle, double xShift, double yShift, int interpolation)
{
    static const int npp_inter[] = {NPPI_INTER_NN, NPPI_INTER_LINEAR, NPPI_INTER_CUBIC};

    CV_Assert(src.type() == CV_8UC1 || src.type() == CV_8UC4);
    CV_Assert(interpolation == INTER_NEAREST || interpolation == INTER_LINEAR || interpolation == INTER_CUBIC);

    dst.create(dsize, src.type());

    NppiSize srcsz;
    srcsz.height = src.rows;
    srcsz.width = src.cols;
    NppiRect srcroi;
    srcroi.x = srcroi.y = 0;
    srcroi.height = src.rows;
    srcroi.width = src.cols;
    NppiRect dstroi;
    dstroi.x = dstroi.y = 0;
    dstroi.height = dst.rows;
    dstroi.width = dst.cols;

    if (src.type() == CV_8UC1)
    {
        nppSafeCall( nppiRotate_8u_C1R(src.ptr<Npp8u>(), srcsz, src.step, srcroi,
            dst.ptr<Npp8u>(), dst.step, dstroi, angle, xShift, yShift, npp_inter[interpolation]) );
    }
    else
    {
        nppSafeCall( nppiRotate_8u_C4R(src.ptr<Npp8u>(), srcsz, src.step, srcroi,
            dst.ptr<Npp8u>(), dst.step, dstroi, angle, xShift, yShift, npp_inter[interpolation]) );
    }
}

////////////////////////////////////////////////////////////////////////
// integral

void cv::gpu::integral(GpuMat& src, GpuMat& sum, GpuMat& sqsum)
{
    CV_Assert(src.type() == CV_8UC1);

    int w = src.cols + 1, h = src.rows + 1;

    sum.create(h, w, CV_32S);
    sqsum.create(h, w, CV_32F);

    NppiSize sz;
    sz.width = src.cols;
    sz.height = src.rows;

    nppSafeCall( nppiSqrIntegral_8u32s32f_C1R(src.ptr<Npp8u>(), src.step, sum.ptr<Npp32s>(),
        sum.step, sqsum.ptr<Npp32f>(), sqsum.step, sz, 0, 0.0f, h) );
}

void cv::gpu::rectStdDev(const GpuMat& src, const GpuMat& sqr, GpuMat& dst, const Rect& rect)
{
    CV_Assert(src.type() == CV_32SC1 && sqr.type() == CV_32FC1);

    dst.create(src.size(), CV_32FC1);

    NppiSize sz;
    sz.width = src.cols;
    sz.height = src.rows;

    NppiRect nppRect;
    nppRect.height = rect.height;
    nppRect.width = rect.width;
    nppRect.x = rect.x;
    nppRect.y = rect.y;

    nppSafeCall( nppiRectStdDev_32s32f_C1R(src.ptr<Npp32s>(), src.step, sqr.ptr<Npp32f>(), sqr.step,
                dst.ptr<Npp32f>(), dst.step, sz, nppRect) );
}

////////////////////////////////////////////////////////////////////////
// Canny

void cv::gpu::Canny(const GpuMat& image, GpuMat& edges, double threshold1, double threshold2, int apertureSize)
{
#if NPP_VERSION >= 32
    CV_Assert(!"disabled until fix crash");
    CV_Assert(image.type() == CV_8UC1);

    GpuMat srcDx, srcDy;

    Sobel(image, srcDx, -1, 1, 0, apertureSize);
    Sobel(image, srcDy, -1, 0, 1, apertureSize);

    srcDx.convertTo(srcDx, CV_32F);
    srcDy.convertTo(srcDy, CV_32F);

    edges.create(image.size(), CV_8UC1);

    NppiSize sz;
    sz.height = image.rows;
    sz.width = image.cols;

    int bufsz;
    nppSafeCall( nppiCannyGetBufferSize(sz, &bufsz) );
    GpuMat buf(1, bufsz, CV_8UC1);

    nppSafeCall( nppiCanny_32f8u_C1R(srcDx.ptr<Npp32f>(), srcDx.step, srcDy.ptr<Npp32f>(), srcDy.step,
        edges.ptr<Npp8u>(), edges.step, sz, (Npp32f)threshold1, (Npp32f)threshold2, buf.ptr<Npp8u>()) );
#else
    CV_Assert(!"This function doesn't supported");
#endif
}

////////////////////////////////////////////////////////////////////////
// Histogram

namespace
{
    template<int n> struct NPPTypeTraits;
    template<> struct NPPTypeTraits<CV_8U>  { typedef Npp8u npp_type; };
    template<> struct NPPTypeTraits<CV_16U> { typedef Npp16u npp_type; };
    template<> struct NPPTypeTraits<CV_16S> { typedef Npp16s npp_type; };
    template<> struct NPPTypeTraits<CV_32F> { typedef Npp32f npp_type; };

    typedef NppStatus (*get_buf_size_c1_t)(NppiSize oSizeROI, int nLevels, int* hpBufferSize);
    typedef NppStatus (*get_buf_size_c4_t)(NppiSize oSizeROI, int nLevels[], int* hpBufferSize);

    template<int SDEPTH> struct NppHistogramEvenFuncC1
    {
        typedef typename NPPTypeTraits<SDEPTH>::npp_type src_t;

	typedef NppStatus (*func_ptr)(const src_t* pSrc, int nSrcStep, NppiSize oSizeROI, Npp32s * pHist,
		    int nLevels, Npp32s nLowerLevel, Npp32s nUpperLevel, Npp8u * pBuffer);
    };
    template<int SDEPTH> struct NppHistogramEvenFuncC4
    {
        typedef typename NPPTypeTraits<SDEPTH>::npp_type src_t;

        typedef NppStatus (*func_ptr)(const src_t* pSrc, int nSrcStep, NppiSize oSizeROI,
            Npp32s * pHist[4], int nLevels[4], Npp32s nLowerLevel[4], Npp32s nUpperLevel[4], Npp8u * pBuffer);
    };

    template<int SDEPTH, typename NppHistogramEvenFuncC1<SDEPTH>::func_ptr func, get_buf_size_c1_t get_buf_size>
    struct NppHistogramEvenC1
    {
        typedef typename NppHistogramEvenFuncC1<SDEPTH>::src_t src_t;

        static void hist(const GpuMat& src, GpuMat& hist, int histSize, int lowerLevel, int upperLevel)
        {
            int levels = histSize + 1;
            hist.create(1, histSize, CV_32S);

            NppiSize sz;
            sz.width = src.cols;
            sz.height = src.rows;

            GpuMat buffer;
            int buf_size;

            get_buf_size(sz, levels, &buf_size);
            buffer.create(1, buf_size, CV_8U);
            nppSafeCall( func(src.ptr<src_t>(), src.step, sz, hist.ptr<Npp32s>(), levels,
                lowerLevel, upperLevel, buffer.ptr<Npp8u>()) );
        }
    };
    template<int SDEPTH, typename NppHistogramEvenFuncC4<SDEPTH>::func_ptr func, get_buf_size_c4_t get_buf_size>
    struct NppHistogramEvenC4
    {
        typedef typename NppHistogramEvenFuncC4<SDEPTH>::src_t src_t;

        static void hist(const GpuMat& src, GpuMat hist[4], int histSize[4], int lowerLevel[4], int upperLevel[4])
        {
            int levels[] = {histSize[0] + 1, histSize[1] + 1, histSize[2] + 1, histSize[3] + 1};
            hist[0].create(1, histSize[0], CV_32S);
            hist[1].create(1, histSize[1], CV_32S);
            hist[2].create(1, histSize[2], CV_32S);
            hist[3].create(1, histSize[3], CV_32S);

            NppiSize sz;
            sz.width = src.cols;
            sz.height = src.rows;

            Npp32s* pHist[] = {hist[0].ptr<Npp32s>(), hist[1].ptr<Npp32s>(), hist[2].ptr<Npp32s>(), hist[3].ptr<Npp32s>()};

            GpuMat buffer;
            int buf_size;

            get_buf_size(sz, levels, &buf_size);
            buffer.create(1, buf_size, CV_8U);
            nppSafeCall( func(src.ptr<src_t>(), src.step, sz, pHist, levels, lowerLevel, upperLevel, buffer.ptr<Npp8u>()) );
        }
    };

    template<int SDEPTH> struct NppHistogramRangeFuncC1
    {
        typedef typename NPPTypeTraits<SDEPTH>::npp_type src_t;
        typedef Npp32s level_t;
        enum {LEVEL_TYPE_CODE=CV_32SC1};

        typedef NppStatus (*func_ptr)(const src_t* pSrc, int nSrcStep, NppiSize oSizeROI, Npp32s* pHist,
            const Npp32s* pLevels, int nLevels, Npp8u* pBuffer);
    };
    template<> struct NppHistogramRangeFuncC1<CV_32F>
    {
        typedef Npp32f src_t;
        typedef Npp32f level_t;
        enum {LEVEL_TYPE_CODE=CV_32FC1};

        typedef NppStatus (*func_ptr)(const Npp32f* pSrc, int nSrcStep, NppiSize oSizeROI, Npp32s* pHist,
            const Npp32f* pLevels, int nLevels, Npp8u* pBuffer);
    };
    template<int SDEPTH> struct NppHistogramRangeFuncC4
    {
        typedef typename NPPTypeTraits<SDEPTH>::npp_type src_t;
        typedef Npp32s level_t;
        enum {LEVEL_TYPE_CODE=CV_32SC1};

        typedef NppStatus (*func_ptr)(const src_t* pSrc, int nSrcStep, NppiSize oSizeROI, Npp32s* pHist[4],
            const Npp32s* pLevels[4], int nLevels[4], Npp8u* pBuffer);
    };
    template<> struct NppHistogramRangeFuncC4<CV_32F>
    {
        typedef Npp32f src_t;
        typedef Npp32f level_t;
        enum {LEVEL_TYPE_CODE=CV_32FC1};

        typedef NppStatus (*func_ptr)(const Npp32f* pSrc, int nSrcStep, NppiSize oSizeROI, Npp32s* pHist[4],
            const Npp32f* pLevels[4], int nLevels[4], Npp8u* pBuffer);
    };

    template<int SDEPTH, typename NppHistogramRangeFuncC1<SDEPTH>::func_ptr func, get_buf_size_c1_t get_buf_size>
    struct NppHistogramRangeC1
    {
        typedef typename NppHistogramRangeFuncC1<SDEPTH>::src_t src_t;
        typedef typename NppHistogramRangeFuncC1<SDEPTH>::level_t level_t;
        enum {LEVEL_TYPE_CODE=NppHistogramRangeFuncC1<SDEPTH>::LEVEL_TYPE_CODE};

        static void hist(const GpuMat& src, GpuMat& hist, const GpuMat& levels)
        {
            CV_Assert(levels.type() == LEVEL_TYPE_CODE && levels.rows == 1);

            hist.create(1, levels.cols - 1, CV_32S);

            NppiSize sz;
            sz.width = src.cols;
            sz.height = src.rows;

            GpuMat buffer;
            int buf_size;

            get_buf_size(sz, levels.cols, &buf_size);
            buffer.create(1, buf_size, CV_8U);
            nppSafeCall( func(src.ptr<src_t>(), src.step, sz, hist.ptr<Npp32s>(), levels.ptr<level_t>(), levels.cols, buffer.ptr<Npp8u>()) );
        }
    };
    template<int SDEPTH, typename NppHistogramRangeFuncC4<SDEPTH>::func_ptr func, get_buf_size_c4_t get_buf_size>
    struct NppHistogramRangeC4
    {
        typedef typename NppHistogramRangeFuncC4<SDEPTH>::src_t src_t;
        typedef typename NppHistogramRangeFuncC1<SDEPTH>::level_t level_t;
        enum {LEVEL_TYPE_CODE=NppHistogramRangeFuncC1<SDEPTH>::LEVEL_TYPE_CODE};

        static void hist(const GpuMat& src, GpuMat hist[4], const GpuMat levels[4])
        {
            CV_Assert(levels[0].type() == LEVEL_TYPE_CODE && levels[0].rows == 1);
            CV_Assert(levels[1].type() == LEVEL_TYPE_CODE && levels[1].rows == 1);
            CV_Assert(levels[2].type() == LEVEL_TYPE_CODE && levels[2].rows == 1);
            CV_Assert(levels[3].type() == LEVEL_TYPE_CODE && levels[3].rows == 1);

            hist[0].create(1, levels[0].cols - 1, CV_32S);
            hist[1].create(1, levels[1].cols - 1, CV_32S);
            hist[2].create(1, levels[2].cols - 1, CV_32S);
            hist[3].create(1, levels[3].cols - 1, CV_32S);

            Npp32s* pHist[] = {hist[0].ptr<Npp32s>(), hist[1].ptr<Npp32s>(), hist[2].ptr<Npp32s>(), hist[3].ptr<Npp32s>()};
            int nLevels[] = {levels[0].cols, levels[1].cols, levels[2].cols, levels[3].cols};
            const level_t* pLevels[] = {levels[0].ptr<level_t>(), levels[1].ptr<level_t>(), levels[2].ptr<level_t>(), levels[3].ptr<level_t>()};

            NppiSize sz;
            sz.width = src.cols;
            sz.height = src.rows;

            GpuMat buffer;
            int buf_size;

            get_buf_size(sz, nLevels, &buf_size);
            buffer.create(1, buf_size, CV_8U);
            nppSafeCall( func(src.ptr<src_t>(), src.step, sz, pHist, pLevels, nLevels, buffer.ptr<Npp8u>()) );
        }
    };
}

void cv::gpu::evenLevels(GpuMat& levels, int nLevels, int lowerLevel, int upperLevel)
{
#if NPP_VERSION >= 32
    Mat host_levels(1, nLevels, CV_32SC1);
    nppSafeCall( nppiEvenLevelsHost_32s(host_levels.ptr<Npp32s>(), nLevels, lowerLevel, upperLevel) );
    levels.upload(host_levels);
#else
    CV_Assert(!"This function doesn't supported");
#endif
}

void cv::gpu::histEven(const GpuMat& src, GpuMat& hist, int histSize, int lowerLevel, int upperLevel)
{
#if NPP_VERSION >= 32
    CV_Assert(src.type() == CV_8UC1 || src.type() == CV_16UC1 || src.type() == CV_16SC1 );

    typedef void (*hist_t)(const GpuMat& src, GpuMat& hist, int levels, int lowerLevel, int upperLevel);
    static const hist_t hist_callers[] =
    {
        NppHistogramEvenC1<CV_8U , nppiHistogramEven_8u_C1R , nppiHistogramEvenGetBufferSize_8u_C1R >::hist,
        0,
        NppHistogramEvenC1<CV_16U, nppiHistogramEven_16u_C1R, nppiHistogramEvenGetBufferSize_16u_C1R>::hist,
        NppHistogramEvenC1<CV_16S, nppiHistogramEven_16s_C1R, nppiHistogramEvenGetBufferSize_16s_C1R>::hist
    };

    hist_callers[src.depth()](src, hist, histSize, lowerLevel, upperLevel);
#else
    CV_Assert(!"This function doesn't supported");
#endif
}

void cv::gpu::histEven(const GpuMat& src, GpuMat hist[4], int histSize[4], int lowerLevel[4], int upperLevel[4])
{
#if NPP_VERSION >= 32
    CV_Assert(src.type() == CV_8UC4 || src.type() == CV_16UC4 || src.type() == CV_16SC4 );

    typedef void (*hist_t)(const GpuMat& src, GpuMat hist[4], int levels[4], int lowerLevel[4], int upperLevel[4]);
    static const hist_t hist_callers[] =
    {
        NppHistogramEvenC4<CV_8U , nppiHistogramEven_8u_C4R , nppiHistogramEvenGetBufferSize_8u_C4R >::hist,
        0,
        NppHistogramEvenC4<CV_16U, nppiHistogramEven_16u_C4R, nppiHistogramEvenGetBufferSize_16u_C4R>::hist,
        NppHistogramEvenC4<CV_16S, nppiHistogramEven_16s_C4R, nppiHistogramEvenGetBufferSize_16s_C4R>::hist
    };

    hist_callers[src.depth()](src, hist, histSize, lowerLevel, upperLevel);
#else
    CV_Assert(!"This function doesn't supported");
#endif
}

void cv::gpu::histRange(const GpuMat& src, GpuMat& hist, const GpuMat& levels)
{
#if NPP_VERSION >= 32
    CV_Assert(src.type() == CV_8UC1 || src.type() == CV_16UC1 || src.type() == CV_16SC1 || src.type() == CV_32FC1);

    typedef void (*hist_t)(const GpuMat& src, GpuMat& hist, const GpuMat& levels);
    static const hist_t hist_callers[] =
    {
        NppHistogramRangeC1<CV_8U , nppiHistogramRange_8u_C1R , nppiHistogramRangeGetBufferSize_8u_C1R >::hist,
        0,
        NppHistogramRangeC1<CV_16U, nppiHistogramRange_16u_C1R, nppiHistogramRangeGetBufferSize_16u_C1R>::hist,
        NppHistogramRangeC1<CV_16S, nppiHistogramRange_16s_C1R, nppiHistogramRangeGetBufferSize_16s_C1R>::hist,
        0,
        NppHistogramRangeC1<CV_32F, nppiHistogramRange_32f_C1R, nppiHistogramRangeGetBufferSize_32f_C1R>::hist
    };

    hist_callers[src.depth()](src, hist, levels);
#else
    CV_Assert(!"This function doesn't supported");
#endif
}

void cv::gpu::histRange(const GpuMat& src, GpuMat hist[4], const GpuMat levels[4])
{
#if NPP_VERSION >= 32
    CV_Assert(src.type() == CV_8UC4 || src.type() == CV_16UC4 || src.type() == CV_16SC4 || src.type() == CV_32FC4);

    typedef void (*hist_t)(const GpuMat& src, GpuMat hist[4], const GpuMat levels[4]);
    static const hist_t hist_callers[] =
    {
        NppHistogramRangeC4<CV_8U , nppiHistogramRange_8u_C4R , nppiHistogramRangeGetBufferSize_8u_C4R >::hist,
        0,
        NppHistogramRangeC4<CV_16U, nppiHistogramRange_16u_C4R, nppiHistogramRangeGetBufferSize_16u_C4R>::hist,
        NppHistogramRangeC4<CV_16S, nppiHistogramRange_16s_C4R, nppiHistogramRangeGetBufferSize_16s_C4R>::hist,
        0,
        NppHistogramRangeC4<CV_32F, nppiHistogramRange_32f_C4R, nppiHistogramRangeGetBufferSize_32f_C4R>::hist
    };

    hist_callers[src.depth()](src, hist, levels);
#else
    CV_Assert(!"This function doesn't supported");
#endif
}

#endif /* !defined (HAVE_CUDA) */