Fix some typos

doc/tutorials/imgproc/random_generator_and_text/random_generator_and_text.markdown

@@ -210,12 +210,12 @@ Explanation
     @code{.cpp}
     image2 = image - Scalar::all(i)
     @endcode
-    So, **image2** is the substraction of **image** and **Scalar::all(i)**. In fact, what happens
+    So, **image2** is the subtraction of **image** and **Scalar::all(i)**. In fact, what happens
-    here is that every pixel of **image2** will be the result of substracting every pixel of
+    here is that every pixel of **image2** will be the result of subtracting every pixel of
     **image** minus the value of **i** (remember that for each pixel we are considering three values
     such as R, G and B, so each of them will be affected)
 
-    Also remember that the substraction operation *always* performs internally a **saturate**
+    Also remember that the subtraction operation *always* performs internally a **saturate**
     operation, which means that the result obtained will always be inside the allowed range (no
     negative and between 0 and 255 for our example).
 

modules/core/include/opencv2/core/hal/msa_macros.h

@@ -502,7 +502,7 @@ typedef double v1f64 __attribute__ ((vector_size(8), aligned(8)));
   (v4u32)__builtin_msa_pckev_w((v4i32)__builtin_msa_sat_u_d((v2u64)__e, 31), (v4i32)__builtin_msa_sat_u_d((v2u64)__d, 31)); \
 })
 
-/* Minimum values between corresponding elements in the two vectors are written to teh returned vector. */
+/* Minimum values between corresponding elements in the two vectors are written to the returned vector. */
 #define msa_minq_s8(__a, __b)  (__builtin_msa_min_s_b(__a, __b))
 #define msa_minq_s16(__a, __b) (__builtin_msa_min_s_h(__a, __b))
 #define msa_minq_s32(__a, __b) (__builtin_msa_min_s_w(__a, __b))
@@ -514,7 +514,7 @@ typedef double v1f64 __attribute__ ((vector_size(8), aligned(8)));
 #define msa_minq_f32(__a, __b) (__builtin_msa_fmin_w(__a, __b))
 #define msa_minq_f64(__a, __b) (__builtin_msa_fmin_d(__a, __b))
 
-/* Maximum values between corresponding elements in the two vectors are written to teh returned vector. */
+/* Maximum values between corresponding elements in the two vectors are written to the returned vector. */
 #define msa_maxq_s8(__a, __b)  (__builtin_msa_max_s_b(__a, __b))
 #define msa_maxq_s16(__a, __b) (__builtin_msa_max_s_h(__a, __b))
 #define msa_maxq_s32(__a, __b) (__builtin_msa_max_s_w(__a, __b))

modules/dnn/src/cuda/concat.cu

@@ -82,7 +82,7 @@ namespace cv { namespace dnn { namespace cuda4dnn { namespace kernels {
     {
         CV_Assert(is_fully_aligned<T>(output, N));
         CV_Assert(is_fully_aligned<T>(input, N));
-        /* more assertions are required to fully check for vectorization possiblity; check concat() */
+        /* more assertions are required to fully check for vectorization possibility; check concat() */
 
         auto kernel = raw::concat_vec<T, N>;
         auto policy = make_policy(kernel, input.size() / N, 0, stream);
@@ -168,7 +168,7 @@ namespace cv { namespace dnn { namespace cuda4dnn { namespace kernels {
         CV_Assert(output.rank() == input.rank());
         CV_Assert(output.rank() == offsets.size());
 
-        /* squeezable axes at the begining of both tensors can be eliminated
+        /* squeezable axes at the beginning of both tensors can be eliminated
          *
          * Reasoning:
          * ----------

modules/dnn/src/cuda/padding.cu

@@ -103,7 +103,7 @@ namespace cv { namespace dnn { namespace cuda4dnn { namespace kernels {
         CV_Assert(output.rank() == input.rank());
         CV_Assert(output.rank() == ranges.size());
 
-        /* squeezable axes at the begining of both tensors can be eliminated
+        /* squeezable axes at the beginning of both tensors can be eliminated
          *
          * Reasoning:
          * ----------

modules/dnn/src/cuda/permute.cu

@@ -83,7 +83,7 @@ namespace cv { namespace dnn { namespace cuda4dnn { namespace kernels {
         CV_Assert(input.rank() == order.size());
         CV_Assert(input.size() == output.size());
 
-        /* squeezable axes at the begining of both tensors which aren't permuted can be eliminated
+        /* squeezable axes at the beginning of both tensors which aren't permuted can be eliminated
          *
          * Reasoning:
          * ----------

modules/dnn/src/cuda/slice.cu

@@ -79,7 +79,7 @@ namespace cv { namespace dnn { namespace cuda4dnn { namespace kernels {
         CV_Assert(output.rank() == input.rank());
         CV_Assert(output.rank() == offsets.size());
 
-        /* squeezable axes at the begining of both tensors can be eliminated
+        /* squeezable axes at the beginning of both tensors can be eliminated
          *
          * Reasoning:
          * ----------

modules/dnn/src/cuda4dnn/csl/cudnn/cudnn.hpp

@@ -218,7 +218,7 @@ namespace cv { namespace dnn { namespace cuda4dnn { namespace csl { namespace cu
                      *
                      * cuDNN frequently assumes that the first axis is the batch axis and the
                      * second axis is the channel axis; hence, we copy the shape of a lower rank
-                     * tensor to the begining of `dims`
+                     * tensor to the beginning of `dims`
                      */
                     std::copy(start, end, std::begin(dims));
 

modules/dnn/src/cuda4dnn/csl/tensor.hpp

@@ -53,7 +53,7 @@ namespace cv { namespace dnn { namespace cuda4dnn { namespace csl {
      * "TensorType" is used when only meta-information such as the size or shape is required, i.e. the data won't be touched
      */
 
-    /** if the \p axis is a negative index, the equivalent postive index is returned; otherwise, returns \p axis */
+    /** if the \p axis is a negative index, the equivalent positive index is returned; otherwise, returns \p axis */
     CUDA4DNN_HOST_DEVICE constexpr std::size_t clamp_axis(int axis, std::size_t rank) {
         return axis < 0 ? axis + rank : axis;
     }

modules/dnn/src/cuda4dnn/primitives/normalize_bbox.hpp

@@ -41,7 +41,7 @@ namespace cv { namespace dnn { namespace cuda4dnn {
         /* 1 for L1 norm, 2 for L2 norm */
         std::size_t norm;
 
-        /* epsilon to use to avoid divison by zero */
+        /* epsilon to use to avoid division by zero */
         T eps;
     };
 

modules/dnn/src/cuda4dnn/primitives/pooling.hpp

@@ -168,7 +168,7 @@ namespace cv { namespace dnn { namespace cuda4dnn {
                  * copying the input to a bigger tensor and padding the ends manually
                  *
                  * But we first try to avoid the transformation using cuDNN's flexibility. cuDNN can accept a smaller or
-                 * a bigger output shape. This effectively allows us to have arbitary padding at the right.
+                 * a bigger output shape. This effectively allows us to have arbitrary padding at the right.
                  */
                 if (std::any_of(std::begin(padding_left), std::end(padding_left), is_not_zero))
                 {

modules/dnn/src/op_cuda.hpp

@@ -65,7 +65,7 @@ namespace cv { namespace dnn {
          * \param[out]  destTensor  destination tensor
          * \param       stream      CUDA stream to use for the memory transfer
          *
-         * The memory copy starts from begining \p srcMat. The number of elements copied is
+         * The memory copy starts from beginning \p srcMat. The number of elements copied is
          * equal to the number of elements in \p destTensor.
          *
          * Pre-conditions:

modules/gapi/test/common/gapi_core_tests.hpp

@@ -73,7 +73,7 @@ inline std::ostream& operator<<(std::ostream& os, bitwiseOp op)
 // 1. Default parameters: int type, cv::Size sz, int dtype, getCompileArgs() function
 //      - available in test body
 // 2. Input/output matrices will be initialized by initMatsRandU (in this fixture)
-// 3. Specific parameters: opType, testWithScalar, scale, doReverseOp of correponding types
+// 3. Specific parameters: opType, testWithScalar, scale, doReverseOp of corresponding types
 //      - created (and initialized) automatically
 //      - available in test body
 // Note: all parameter _values_ (e.g. type CV_8UC3) are set via INSTANTIATE_TEST_CASE_P macro

modules/gapi/test/common/gapi_imgproc_tests.hpp

@@ -25,7 +25,7 @@ namespace opencv_test
 // 1. Default parameters: int type, cv::Size sz, int dtype, getCompileArgs() function
 //      - available in test body
 // 2. Input/output matrices will be initialized by initMatrixRandN (in this fixture)
-// 3. Specific parameters: cmpF, kernSize, borderType of correponding types
+// 3. Specific parameters: cmpF, kernSize, borderType of corresponding types
 //      - created (and initialized) automatically
 //      - available in test body
 // Note: all parameter _values_ (e.g. type CV_8UC3) are set via INSTANTIATE_TEST_CASE_P macro

modules/gapi/test/common/gapi_operators_tests.hpp

@@ -195,7 +195,7 @@ g_api_ocv_pair_mat_mat opXor = {std::string{"operator^"},
 // 1. Default parameters: int type, cv::Size sz, int dtype, getCompileArgs() function
 //      - available in test body
 // 2. Input/output matrices will be initialized by initMatsRandU (in this fixture)
-// 3. Specific parameters: cmpF, op of correponding types
+// 3. Specific parameters: cmpF, op of corresponding types
 //      - created (and initialized) automatically
 //      - available in test body
 // Note: all parameter _values_ (e.g. type CV_8UC3) are set via INSTANTIATE_TEST_CASE_P macro

modules/gapi/test/streaming/gapi_streaming_tests.cpp

@@ -540,7 +540,7 @@ TEST(GAPI_Streaming_Types, XChangeVector)
     auto fluid_kernels = cv::gapi::core::fluid::kernels();
     fluid_kernels.include<TypesTest::FluidAddV>();
 
-    // Here OCV takes precedense over Fluid, whith SubC & SumV remaining
+    // Here OCV takes precedense over Fluid, with SubC & SumV remaining
     // in Fluid.
     auto kernels = cv::gapi::combine(fluid_kernels, ocv_kernels);
 

modules/imgproc/test/test_watershed.cpp

@@ -83,7 +83,7 @@ void CV_WatershedTest::run( int /* start_from */)
         Point* p = (Point*)cvGetSeqElem(cnts, 0);
 
         //expected image was added with 1 in order to save to png
-        //so now we substract 1 to get real color
+        //so now we subtract 1 to get real color
         if(!exp.empty())
             colors.push_back(exp.ptr(p->y)[p->x] - 1);
     }