Merge pull request #16932 from alalek:imgproc_resize_nn_cleanup

modules/imgproc/src/resize.avx2.cpp

@@ -59,8 +59,8 @@ class resizeNNInvokerAVX4 CV_FINAL :
     public ParallelLoopBody
 {
 public:
-    resizeNNInvokerAVX4(const Mat& _src, Mat &_dst, int *_x_ofs, int _pix_size4, double _ify) :
-        ParallelLoopBody(), src(_src), dst(_dst), x_ofs(_x_ofs), pix_size4(_pix_size4),
+    resizeNNInvokerAVX4(const Mat& _src, Mat &_dst, int *_x_ofs, double _ify) :
+        ParallelLoopBody(), src(_src), dst(_dst), x_ofs(_x_ofs),
         ify(_ify)
     {
     }
@@ -129,9 +129,9 @@ public:
     }
 
 private:
-    const Mat src;
-    Mat dst;
-    int* x_ofs, pix_size4;
+    const Mat& src;
+    Mat& dst;
+    int* x_ofs;
     double ify;
 
     resizeNNInvokerAVX4(const resizeNNInvokerAVX4&);
@@ -142,8 +142,8 @@ class resizeNNInvokerAVX2 CV_FINAL :
     public ParallelLoopBody
 {
 public:
-    resizeNNInvokerAVX2(const Mat& _src, Mat &_dst, int *_x_ofs, int _pix_size4, double _ify) :
-        ParallelLoopBody(), src(_src), dst(_dst), x_ofs(_x_ofs), pix_size4(_pix_size4),
+    resizeNNInvokerAVX2(const Mat& _src, Mat &_dst, int *_x_ofs, double _ify) :
+        ParallelLoopBody(), src(_src), dst(_dst), x_ofs(_x_ofs),
         ify(_ify)
     {
     }
@@ -235,24 +235,24 @@ public:
     }
 
 private:
-    const Mat src;
-    Mat dst;
-    int* x_ofs, pix_size4;
+    const Mat& src;
+    Mat& dst;
+    int* x_ofs;
     double ify;
 
     resizeNNInvokerAVX2(const resizeNNInvokerAVX2&);
     resizeNNInvokerAVX2& operator=(const resizeNNInvokerAVX2&);
 };
 
-void resizeNN2_AVX2(const Range& range, const Mat& src, Mat &dst, int *x_ofs, int pix_size4, double ify)
+void resizeNN2_AVX2(const Range& range, const Mat& src, Mat &dst, int *x_ofs, double ify)
 {
-    resizeNNInvokerAVX2 invoker(src, dst, x_ofs, pix_size4, ify);
+    resizeNNInvokerAVX2 invoker(src, dst, x_ofs, ify);
     parallel_for_(range, invoker, dst.total() / (double)(1 << 16));
 }
 
-void resizeNN4_AVX2(const Range& range, const Mat& src, Mat &dst, int *x_ofs, int pix_size4, double ify)
+void resizeNN4_AVX2(const Range& range, const Mat& src, Mat &dst, int *x_ofs, double ify)
 {
-    resizeNNInvokerAVX4 invoker(src, dst, x_ofs, pix_size4, ify);
+    resizeNNInvokerAVX4 invoker(src, dst, x_ofs, ify);
     parallel_for_(range, invoker, dst.total() / (double)(1 << 16));
 }
 

modules/imgproc/src/resize.cpp

@@ -970,8 +970,8 @@ class resizeNNInvoker :
     public ParallelLoopBody
 {
 public:
-    resizeNNInvoker(const Mat& _src, Mat &_dst, int *_x_ofs, int _pix_size4, double _ify) :
-        ParallelLoopBody(), src(_src), dst(_dst), x_ofs(_x_ofs), pix_size4(_pix_size4),
+    resizeNNInvoker(const Mat& _src, Mat &_dst, int *_x_ofs, double _ify) :
+        ParallelLoopBody(), src(_src), dst(_dst), x_ofs(_x_ofs),
         ify(_ify)
     {
     }
@@ -1052,9 +1052,9 @@ public:
     }
 
 private:
-    const Mat src;
-    Mat dst;
-    int* x_ofs, pix_size4;
+    const Mat& src;
+    Mat& dst;
+    int* x_ofs;
     double ify;
 
     resizeNNInvoker(const resizeNNInvoker&);
@@ -1068,7 +1068,6 @@ resizeNN( const Mat& src, Mat& dst, double fx, double fy )
     AutoBuffer<int> _x_ofs(dsize.width);
     int* x_ofs = _x_ofs.data();
     int pix_size = (int)src.elemSize();
-    int pix_size4 = (int)(pix_size / sizeof(int));
     double ifx = 1./fx, ify = 1./fy;
     int x;
 
@@ -1083,9 +1082,9 @@ resizeNN( const Mat& src, Mat& dst, double fx, double fy )
     if(CV_CPU_HAS_SUPPORT_AVX2 && ((pix_size == 2) || (pix_size == 4)))
     {
         if(pix_size == 2)
-            opt_AVX2::resizeNN2_AVX2(range, src, dst, x_ofs, pix_size4, ify);
+            opt_AVX2::resizeNN2_AVX2(range, src, dst, x_ofs, ify);
         else
-            opt_AVX2::resizeNN4_AVX2(range, src, dst, x_ofs, pix_size4, ify);
+            opt_AVX2::resizeNN4_AVX2(range, src, dst, x_ofs, ify);
     }
     else
 #endif
@@ -1093,14 +1092,14 @@ resizeNN( const Mat& src, Mat& dst, double fx, double fy )
     if(CV_CPU_HAS_SUPPORT_SSE4_1 && ((pix_size == 2) || (pix_size == 4)))
     {
         if(pix_size == 2)
-            opt_SSE4_1::resizeNN2_SSE4_1(range, src, dst, x_ofs, pix_size4, ify);
+            opt_SSE4_1::resizeNN2_SSE4_1(range, src, dst, x_ofs, ify);
         else
-            opt_SSE4_1::resizeNN4_SSE4_1(range, src, dst, x_ofs, pix_size4, ify);
+            opt_SSE4_1::resizeNN4_SSE4_1(range, src, dst, x_ofs, ify);
     }
     else
 #endif
     {
-        resizeNNInvoker invoker(src, dst, x_ofs, pix_size4, ify);
+        resizeNNInvoker invoker(src, dst, x_ofs, ify);
         parallel_for_(range, invoker, dst.total()/(double)(1<<16));
     }
 }

modules/imgproc/src/resize.hpp

@@ -56,16 +56,16 @@ namespace cv
 namespace opt_AVX2
 {
 #if CV_TRY_AVX2
-void resizeNN2_AVX2(const Range&, const Mat&, Mat&, int*, int, double);
-void resizeNN4_AVX2(const Range&, const Mat&, Mat&, int*, int, double);
+void resizeNN2_AVX2(const Range&, const Mat&, Mat&, int*, double);
+void resizeNN4_AVX2(const Range&, const Mat&, Mat&, int*, double);
 #endif
 }
 
 namespace opt_SSE4_1
 {
 #if CV_TRY_SSE4_1
-void resizeNN2_SSE4_1(const Range&, const Mat&, Mat&, int*, int, double);
-void resizeNN4_SSE4_1(const Range&, const Mat&, Mat&, int*, int, double);
+void resizeNN2_SSE4_1(const Range&, const Mat&, Mat&, int*, double);
+void resizeNN4_SSE4_1(const Range&, const Mat&, Mat&, int*, double);
 
 int VResizeLanczos4Vec_32f16u_SSE41(const float** src, ushort* dst, const float* beta, int width);
 #endif

modules/imgproc/src/resize.sse4_1.cpp

@@ -59,8 +59,8 @@ class resizeNNInvokerSSE2 :
     public ParallelLoopBody
 {
 public:
-    resizeNNInvokerSSE2(const Mat& _src, Mat &_dst, int *_x_ofs, int _pix_size4, double _ify) :
-        ParallelLoopBody(), src(_src), dst(_dst), x_ofs(_x_ofs), pix_size4(_pix_size4),
+    resizeNNInvokerSSE2(const Mat& _src, Mat &_dst, int *_x_ofs, double _ify) :
+        ParallelLoopBody(), src(_src), dst(_dst), x_ofs(_x_ofs),
         ify(_ify)
     {
     }
@@ -110,9 +110,9 @@ public:
     }
 
 private:
-    const Mat src;
-    Mat dst;
-    int* x_ofs, pix_size4;
+    const Mat& src;
+    Mat& dst;
+    int* x_ofs;
     double ify;
 
     resizeNNInvokerSSE2(const resizeNNInvokerSSE2&);
@@ -123,8 +123,8 @@ class resizeNNInvokerSSE4 :
     public ParallelLoopBody
 {
 public:
-    resizeNNInvokerSSE4(const Mat& _src, Mat &_dst, int *_x_ofs, int _pix_size4, double _ify) :
-        ParallelLoopBody(), src(_src), dst(_dst), x_ofs(_x_ofs), pix_size4(_pix_size4),
+    resizeNNInvokerSSE4(const Mat& _src, Mat &_dst, int *_x_ofs, double _ify) :
+        ParallelLoopBody(), src(_src), dst(_dst), x_ofs(_x_ofs),
         ify(_ify)
     {
     }
@@ -165,24 +165,24 @@ public:
     }
 
 private:
-    const Mat src;
-    Mat dst;
-    int* x_ofs, pix_size4;
+    const Mat& src;
+    Mat& dst;
+    int* x_ofs;
     double ify;
 
     resizeNNInvokerSSE4(const resizeNNInvokerSSE4&);
     resizeNNInvokerSSE4& operator=(const resizeNNInvokerSSE4&);
 };
 
-void resizeNN2_SSE4_1(const Range& range, const Mat& src, Mat &dst, int *x_ofs, int pix_size4, double ify)
+void resizeNN2_SSE4_1(const Range& range, const Mat& src, Mat &dst, int *x_ofs, double ify)
 {
-    resizeNNInvokerSSE2 invoker(src, dst, x_ofs, pix_size4, ify);
+    resizeNNInvokerSSE2 invoker(src, dst, x_ofs, ify);
     parallel_for_(range, invoker, dst.total() / (double)(1 << 16));
 }
 
-void resizeNN4_SSE4_1(const Range& range, const Mat& src, Mat &dst, int *x_ofs, int pix_size4, double ify)
+void resizeNN4_SSE4_1(const Range& range, const Mat& src, Mat &dst, int *x_ofs, double ify)
 {
-    resizeNNInvokerSSE4 invoker(src, dst, x_ofs, pix_size4, ify);
+    resizeNNInvokerSSE4 invoker(src, dst, x_ofs, ify);
     parallel_for_(range, invoker, dst.total() / (double)(1 << 16));
 }