graphcuts.cpp 11.5 KB
Newer Older
wester committed
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
/*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"

wester committed
45 46
// GraphCut has been removed in NPP 8.0
#if !defined (HAVE_CUDA) || defined (CUDA_DISABLER) || (CUDART_VERSION >= 8000)
wester committed
47

wester committed
48 49
void cv::gpu::graphcut(GpuMat&, GpuMat&, GpuMat&, GpuMat&, GpuMat&, GpuMat&, GpuMat&, Stream&) { throw_nogpu(); }
void cv::gpu::graphcut(GpuMat&, GpuMat&, GpuMat&, GpuMat&, GpuMat&, GpuMat&, GpuMat&, GpuMat&, GpuMat&, GpuMat&, GpuMat&, Stream&) { throw_nogpu(); }
wester committed
50

wester committed
51 52
void cv::gpu::connectivityMask(const GpuMat&, GpuMat&, const cv::Scalar&, const cv::Scalar&, Stream&) { throw_nogpu(); }
void cv::gpu::labelComponents(const GpuMat&, GpuMat&, int, Stream&) { throw_nogpu(); }
wester committed
53 54 55

#else /* !defined (HAVE_CUDA) */

wester committed
56
namespace cv { namespace gpu { namespace device
wester committed
57 58 59 60 61 62 63 64 65 66 67 68 69 70 71
{
    namespace ccl
    {
        void labelComponents(const PtrStepSzb& edges, PtrStepSzi comps, int flags, cudaStream_t stream);

        template<typename T>
        void computeEdges(const PtrStepSzb& image, PtrStepSzb edges, const float4& lo, const float4& hi, cudaStream_t stream);
    }
}}}

static float4 scalarToCudaType(const cv::Scalar& in)
{
  return make_float4((float)in[0], (float)in[1], (float)in[2], (float)in[3]);
}

wester committed
72
void cv::gpu::connectivityMask(const GpuMat& image, GpuMat& mask, const cv::Scalar& lo, const cv::Scalar& hi, Stream& s)
wester committed
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
{
    CV_Assert(!image.empty());

    int ch = image.channels();
    CV_Assert(ch <= 4);

    int depth = image.depth();

    typedef void (*func_t)(const PtrStepSzb& image, PtrStepSzb edges, const float4& lo, const float4& hi, cudaStream_t stream);

    static const func_t suppotLookup[8][4] =
    {   //    1,    2,     3,     4
        { device::ccl::computeEdges<uchar>,  0,  device::ccl::computeEdges<uchar3>,  device::ccl::computeEdges<uchar4>  },// CV_8U
        { 0,                                 0,  0,                                  0                                  },// CV_16U
        { device::ccl::computeEdges<ushort>, 0,  device::ccl::computeEdges<ushort3>, device::ccl::computeEdges<ushort4> },// CV_8S
        { 0,                                 0,  0,                                  0                                  },// CV_16S
        { device::ccl::computeEdges<int>,    0,  0,                                  0                                  },// CV_32S
        { device::ccl::computeEdges<float>,  0,  0,                                  0                                  },// CV_32F
        { 0,                                 0,  0,                                  0                                  },// CV_64F
        { 0,                                 0,  0,                                  0                                  } // CV_USRTYPE1
    };

    func_t f = suppotLookup[depth][ch - 1];
    CV_Assert(f);

    if (image.size() != mask.size() || mask.type() != CV_8UC1)
        mask.create(image.size(), CV_8UC1);

    cudaStream_t stream = StreamAccessor::getStream(s);
    float4 culo = scalarToCudaType(lo), cuhi = scalarToCudaType(hi);
    f(image, mask, culo, cuhi, stream);
}

wester committed
106
void cv::gpu::labelComponents(const GpuMat& mask, GpuMat& components, int flags, Stream& s)
wester committed
107 108 109 110
{
    CV_Assert(!mask.empty() && mask.type() == CV_8U);

    if (!deviceSupports(SHARED_ATOMICS))
wester committed
111
        CV_Error(CV_StsNotImplemented, "The device doesn't support shared atomics and communicative synchronization!");
wester committed
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

    components.create(mask.size(), CV_32SC1);

    cudaStream_t stream = StreamAccessor::getStream(s);
    device::ccl::labelComponents(mask, components, flags, stream);
}

namespace
{
    typedef NppStatus (*init_func_t)(NppiSize oSize, NppiGraphcutState** ppState, Npp8u* pDeviceMem);

    class NppiGraphcutStateHandler
    {
    public:
        NppiGraphcutStateHandler(NppiSize sznpp, Npp8u* pDeviceMem, const init_func_t func)
        {
            nppSafeCall( func(sznpp, &pState, pDeviceMem) );
        }

        ~NppiGraphcutStateHandler()
        {
            nppSafeCall( nppiGraphcutFree(pState) );
        }

        operator NppiGraphcutState*()
        {
            return pState;
        }

    private:
        NppiGraphcutState* pState;
    };
}

wester committed
146
void cv::gpu::graphcut(GpuMat& terminals, GpuMat& leftTransp, GpuMat& rightTransp, GpuMat& top, GpuMat& bottom, GpuMat& labels, GpuMat& buf, Stream& s)
wester committed
147
{
wester committed
148
#if (CUDART_VERSION < 5000)
wester committed
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
    CV_Assert(terminals.type() == CV_32S);
#else
    CV_Assert(terminals.type() == CV_32S || terminals.type() == CV_32F);
#endif

    Size src_size = terminals.size();

    CV_Assert(leftTransp.size() == Size(src_size.height, src_size.width));
    CV_Assert(leftTransp.type() == terminals.type());

    CV_Assert(rightTransp.size() == Size(src_size.height, src_size.width));
    CV_Assert(rightTransp.type() == terminals.type());

    CV_Assert(top.size() == src_size);
    CV_Assert(top.type() == terminals.type());

    CV_Assert(bottom.size() == src_size);
    CV_Assert(bottom.type() == terminals.type());

    labels.create(src_size, CV_8U);

    NppiSize sznpp;
    sznpp.width = src_size.width;
    sznpp.height = src_size.height;

    int bufsz;
    nppSafeCall( nppiGraphcutGetSize(sznpp, &bufsz) );

    ensureSizeIsEnough(1, bufsz, CV_8U, buf);

    cudaStream_t stream = StreamAccessor::getStream(s);

    NppStreamHandler h(stream);

    NppiGraphcutStateHandler state(sznpp, buf.ptr<Npp8u>(), nppiGraphcutInitAlloc);

wester committed
185
#if (CUDART_VERSION < 5000)
wester committed
186 187 188 189 190 191 192 193 194 195 196 197 198 199 200 201 202 203 204
    nppSafeCall( nppiGraphcut_32s8u(terminals.ptr<Npp32s>(), leftTransp.ptr<Npp32s>(), rightTransp.ptr<Npp32s>(), top.ptr<Npp32s>(), bottom.ptr<Npp32s>(),
        static_cast<int>(terminals.step), static_cast<int>(leftTransp.step), sznpp, labels.ptr<Npp8u>(), static_cast<int>(labels.step), state) );
#else
    if (terminals.type() == CV_32S)
    {
        nppSafeCall( nppiGraphcut_32s8u(terminals.ptr<Npp32s>(), leftTransp.ptr<Npp32s>(), rightTransp.ptr<Npp32s>(), top.ptr<Npp32s>(), bottom.ptr<Npp32s>(),
            static_cast<int>(terminals.step), static_cast<int>(leftTransp.step), sznpp, labels.ptr<Npp8u>(), static_cast<int>(labels.step), state) );
    }
    else
    {
        nppSafeCall( nppiGraphcut_32f8u(terminals.ptr<Npp32f>(), leftTransp.ptr<Npp32f>(), rightTransp.ptr<Npp32f>(), top.ptr<Npp32f>(), bottom.ptr<Npp32f>(),
            static_cast<int>(terminals.step), static_cast<int>(leftTransp.step), sznpp, labels.ptr<Npp8u>(), static_cast<int>(labels.step), state) );
    }
#endif

    if (stream == 0)
        cudaSafeCall( cudaDeviceSynchronize() );
}

wester committed
205
void cv::gpu::graphcut(GpuMat& terminals, GpuMat& leftTransp, GpuMat& rightTransp, GpuMat& top, GpuMat& topLeft, GpuMat& topRight,
wester committed
206 207
              GpuMat& bottom, GpuMat& bottomLeft, GpuMat& bottomRight, GpuMat& labels, GpuMat& buf, Stream& s)
{
wester committed
208
#if (CUDART_VERSION < 5000)
wester committed
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
    CV_Assert(terminals.type() == CV_32S);
#else
    CV_Assert(terminals.type() == CV_32S || terminals.type() == CV_32F);
#endif

    Size src_size = terminals.size();

    CV_Assert(leftTransp.size() == Size(src_size.height, src_size.width));
    CV_Assert(leftTransp.type() == terminals.type());

    CV_Assert(rightTransp.size() == Size(src_size.height, src_size.width));
    CV_Assert(rightTransp.type() == terminals.type());

    CV_Assert(top.size() == src_size);
    CV_Assert(top.type() == terminals.type());

    CV_Assert(topLeft.size() == src_size);
    CV_Assert(topLeft.type() == terminals.type());

    CV_Assert(topRight.size() == src_size);
    CV_Assert(topRight.type() == terminals.type());

    CV_Assert(bottom.size() == src_size);
    CV_Assert(bottom.type() == terminals.type());

    CV_Assert(bottomLeft.size() == src_size);
    CV_Assert(bottomLeft.type() == terminals.type());

    CV_Assert(bottomRight.size() == src_size);
    CV_Assert(bottomRight.type() == terminals.type());

    labels.create(src_size, CV_8U);

    NppiSize sznpp;
    sznpp.width = src_size.width;
    sznpp.height = src_size.height;

    int bufsz;
    nppSafeCall( nppiGraphcut8GetSize(sznpp, &bufsz) );

    ensureSizeIsEnough(1, bufsz, CV_8U, buf);

    cudaStream_t stream = StreamAccessor::getStream(s);

    NppStreamHandler h(stream);

    NppiGraphcutStateHandler state(sznpp, buf.ptr<Npp8u>(), nppiGraphcut8InitAlloc);

wester committed
257
#if (CUDART_VERSION < 5000)
wester committed
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
    nppSafeCall( nppiGraphcut8_32s8u(terminals.ptr<Npp32s>(), leftTransp.ptr<Npp32s>(), rightTransp.ptr<Npp32s>(),
        top.ptr<Npp32s>(), topLeft.ptr<Npp32s>(), topRight.ptr<Npp32s>(),
        bottom.ptr<Npp32s>(), bottomLeft.ptr<Npp32s>(), bottomRight.ptr<Npp32s>(),
        static_cast<int>(terminals.step), static_cast<int>(leftTransp.step), sznpp, labels.ptr<Npp8u>(), static_cast<int>(labels.step), state) );
#else
    if (terminals.type() == CV_32S)
    {
        nppSafeCall( nppiGraphcut8_32s8u(terminals.ptr<Npp32s>(), leftTransp.ptr<Npp32s>(), rightTransp.ptr<Npp32s>(),
            top.ptr<Npp32s>(), topLeft.ptr<Npp32s>(), topRight.ptr<Npp32s>(),
            bottom.ptr<Npp32s>(), bottomLeft.ptr<Npp32s>(), bottomRight.ptr<Npp32s>(),
            static_cast<int>(terminals.step), static_cast<int>(leftTransp.step), sznpp, labels.ptr<Npp8u>(), static_cast<int>(labels.step), state) );
    }
    else
    {
        nppSafeCall( nppiGraphcut8_32f8u(terminals.ptr<Npp32f>(), leftTransp.ptr<Npp32f>(), rightTransp.ptr<Npp32f>(),
            top.ptr<Npp32f>(), topLeft.ptr<Npp32f>(), topRight.ptr<Npp32f>(),
            bottom.ptr<Npp32f>(), bottomLeft.ptr<Npp32f>(), bottomRight.ptr<Npp32f>(),
            static_cast<int>(terminals.step), static_cast<int>(leftTransp.step), sznpp, labels.ptr<Npp8u>(), static_cast<int>(labels.step), state) );
    }
#endif

    if (stream == 0)
        cudaSafeCall( cudaDeviceSynchronize() );
}

#endif /* !defined (HAVE_CUDA) */