Merge pull request #3924 from CodeLinaro:integrateYUV_fastcv_extn

Fastcv extn for integrate YUV(YCbCr) image and bug fix in split extn

Author: asmorkalov

modules/fastcv/include/opencv2/fastcv/arithm.hpp

@@ -65,6 +65,24 @@ CV_EXPORTS_W void gemm(InputArray src1, InputArray src2, OutputArray dst, float
 
 //! @}
 
+//! @addtogroup fastcv
+//! @{
+
+/**
+ * @brief Integral of a YCbCr420 image.
+ *        Note: Input height should be multiple of 2. Input width and stride should be multiple of 16.
+ *              Output stride should be multiple of 8.
+ *              It is optimized for Qualcomm's processors
+ * @param Y Input Y component of 8UC1 YCbCr420 image.
+ * @param CbCr Input CbCr component(interleaved) of 8UC1 YCbCr420 image.
+ * @param IY Output Y integral of CV_32S one channel, size (Y height + 1)*(Y width + 1)
+ * @param ICb Output Cb integral of CV_32S one channel, size (Y height/2 + 1)*(Y width/2 + 1)
+ * @param ICr Output Cr integral of CV_32S one channel, size (Y height/2 + 1)*(Y width/2 + 1)
+ */
+CV_EXPORTS_W void integrateYUV(InputArray Y, InputArray CbCr, OutputArray IY, OutputArray ICb, OutputArray ICr);
+
+//! @}
+
 } // fastcv::
 } // cv::
 

modules/fastcv/perf/perf_arithm.cpp

@@ -0,0 +1,32 @@
+/*
+ * Copyright (c) 2025 Qualcomm Innovation Center, Inc. All rights reserved.
+ * SPDX-License-Identifier: Apache-2.0
+*/
+
+#include "perf_precomp.hpp"
+
+namespace opencv_test {
+
+typedef perf::TestBaseWithParam<tuple<Size, int>> IntegrateYUVPerfTest;
+
+PERF_TEST_P(IntegrateYUVPerfTest, run,
+    ::testing::Combine(::testing::Values(perf::szVGA, perf::sz720p, perf::sz1080p), // image size
+                       ::testing::Values(CV_8U)                                     // image depth
+                      )
+           )
+{
+    cv::Size srcSize = get<0>(GetParam());
+    int depth = get<1>(GetParam());
+
+    cv::Mat Y(srcSize, depth), CbCr(srcSize.height/2, srcSize.width, depth);
+    cv::Mat IY, ICb, ICr;
+    RNG& rng = cv::theRNG();
+    cvtest::randUni(rng, Y, Scalar::all(0), Scalar::all(255));
+    cvtest::randUni(rng, CbCr, Scalar::all(0), Scalar::all(255));
+
+    TEST_CYCLE() cv::fastcv::integrateYUV(Y, CbCr, IY, ICb, ICr);
+
+    SANITY_CHECK_NOTHING();
+}
+
+} // namespace

modules/fastcv/src/arithm.cpp

@@ -180,5 +180,29 @@ void gemm(InputArray _src1, InputArray _src2, OutputArray _dst, float alpha, Inp
     }
 }
 
+void integrateYUV(InputArray _Y, InputArray _CbCr, OutputArray _IY, OutputArray _ICb, OutputArray _ICr)
+{
+    CV_Assert(!_Y.empty() && !_CbCr.empty());
+    CV_Assert(_Y.type() == _CbCr.type() && _Y.type() == CV_8UC1);
+    Mat Y = _Y.getMat();
+    Mat CbCr = _CbCr.getMat();
+    int Ywidth = Y.cols;
+    int Yheight = Y.rows;
+
+    INITIALIZATION_CHECK;
+
+    _IY.create(Yheight + 1, Ywidth + 1, CV_32SC1);
+    _ICb.create(Yheight/2 + 1, Ywidth/2 + 1, CV_32SC1);
+    _ICr.create(Yheight/2 + 1, Ywidth/2 + 1, CV_32SC1);
+
+    Mat IY_ = _IY.getMat();
+    Mat ICb_ = _ICb.getMat();
+    Mat ICr_ = _ICr.getMat();
+
+    fcvIntegrateImageYCbCr420PseudoPlanaru8(Y.data, CbCr.data, Ywidth, Yheight, Y.step[0],
+                                            CbCr.step[0], (uint32_t*)IY_.data, (uint32_t*)ICb_.data, (uint32_t*)ICr_.data,
+                                            IY_.step[0], ICb_.step[0], ICr_.step[0]);
+}
+
 } // fastcv::
 } // cv::

modules/fastcv/src/channel.cpp

@@ -86,7 +86,7 @@ void split(InputArray _src, OutputArrayOfArrays _mv)
 
     CV_Assert(depth == CV_8U && (cn == 2 || cn == 3 || cn == 4));
     CV_Assert(src.dims <= 2);
-
+    _mv.create(cn, 1, depth);
     for( int k = 0; k < cn; k++ )
     {
         _mv.create(src.dims, src.size, depth, k);

modules/fastcv/test/test_arithm.cpp

@@ -79,6 +79,36 @@ TEST_P(ArithmOpTest, accuracy)
     EXPECT_EQ(normL2, 0);
 }
 
+typedef testing::TestWithParam<tuple<Size>> IntegrateYUVTest;
+
+TEST_P(IntegrateYUVTest, accuracy)
+{
+    auto p = GetParam();
+    Size srcSize = std::get<0>(p);
+    int depth = CV_8U;
+
+    cv::Mat Y(srcSize, depth), CbCr(srcSize.height/2, srcSize.width, depth);
+    cv::Mat IY, ICb, ICr;
+    RNG& rng = cv::theRNG();
+    cvtest::randUni(rng, Y, Scalar::all(0), Scalar::all(255));
+    cvtest::randUni(rng, CbCr, Scalar::all(0), Scalar::all(255));
+
+    cv::fastcv::integrateYUV(Y, CbCr, IY, ICb, ICr);
+
+    CbCr = CbCr.reshape(2,0);
+    std::vector<cv::Mat> ref;
+    cv::fastcv::split(CbCr, ref);
+
+    cv::Mat IY_ref, ICb_ref, ICr_ref;
+    cv::integral(Y,IY_ref,CV_32S);
+    cv::integral(ref[0],ICb_ref,CV_32S);
+    cv::integral(ref[1],ICr_ref,CV_32S);
+
+    EXPECT_EQ(IY_ref.at<int>(IY_ref.rows - 1, IY_ref.cols - 1), IY.at<int>(IY.rows - 1, IY.cols - 1));
+    EXPECT_EQ(ICb_ref.at<int>(ICb_ref.rows - 1, ICb_ref.cols - 1), ICb.at<int>(ICb.rows - 1, ICb.cols - 1));
+    EXPECT_EQ(ICr_ref.at<int>(ICr_ref.rows - 1, ICr_ref.cols - 1), ICr.at<int>(ICr.rows - 1, ICr.cols - 1));
+}
+
 INSTANTIATE_TEST_CASE_P(FastCV_Extension, MatMulTest,
                          ::testing::Combine(::testing::Values(8, 16, 128, 256),   // rows1
                                             ::testing::Values(8, 16, 128, 256),   // cols1
@@ -89,4 +119,7 @@ INSTANTIATE_TEST_CASE_P(FastCV_Extension, ArithmOpTest,
                                             ::testing::Values(CV_8U, CV_16S), // depth
                                             ::testing::Values(0,1))); // op type
 
+INSTANTIATE_TEST_CASE_P(FastCV_Extension, IntegrateYUVTest,
+                         Values(perf::szVGA, perf::sz720p, perf::sz1080p)); // sz
+
 }} // namespaces opencv_test, ::