diff --git a/modules/ocl/perf/perf_imgproc.cpp b/modules/ocl/perf/perf_imgproc.cpp
index a6859d09d4..5eb32b46c9 100644
--- a/modules/ocl/perf/perf_imgproc.cpp
+++ b/modules/ocl/perf/perf_imgproc.cpp
@@ -198,7 +198,7 @@ PERF_TEST_P(cornerHarrisFixture, cornerHarris,
 
 typedef TestBaseWithParam<Size> integralFixture;
 
-PERF_TEST_P(integralFixture, DISABLED_integral, OCL_TYPICAL_MAT_SIZES) // TODO does not work properly
+PERF_TEST_P(integralFixture, integral, OCL_TYPICAL_MAT_SIZES)
 {
     const Size srcSize = GetParam();
 
diff --git a/modules/ocl/src/haar.cpp b/modules/ocl/src/haar.cpp
index aac3785e79..05a76aa84d 100644
--- a/modules/ocl/src/haar.cpp
+++ b/modules/ocl/src/haar.cpp
@@ -1141,7 +1141,6 @@ void cv::ocl::OclCascadeClassifierBuf::detectMultiScale(oclMat &gimg, CV_OUT std
         CvSize sz;
 
         cv::Rect roi, roi2;
-        cv::Mat imgroi, imgroisq;
         cv::ocl::oclMat resizeroi, gimgroi, gimgroisq;
 
         for( int i = 0; i < m_loopcount; i++ )
diff --git a/modules/ocl/src/imgproc.cpp b/modules/ocl/src/imgproc.cpp
index 5e0f54fab5..ff3d95fdae 100644
--- a/modules/ocl/src/imgproc.cpp
+++ b/modules/ocl/src/imgproc.cpp
@@ -975,10 +975,12 @@ namespace cv
         void integral(const oclMat &src, oclMat &sum, oclMat &sqsum)
         {
             CV_Assert(src.type() == CV_8UC1);
-            if(!src.clCxt->supportsFeature(FEATURE_CL_DOUBLE) && src.depth() == CV_64F)
+            if(!src.clCxt->supportsFeature(ocl::FEATURE_CL_DOUBLE) && src.depth() == CV_64F)
             {
                 CV_Error(CV_GpuNotSupported, "select device don't support double");
+                return;
             }
+
             int vlen = 4;
             int offset = src.offset / vlen;
             int pre_invalid = src.offset % vlen;
@@ -986,50 +988,45 @@ namespace cv
 
             oclMat t_sum , t_sqsum;
             int w = src.cols + 1, h = src.rows + 1;
-            int depth;
-            if( src.cols * src.rows <= 2901 * 2901 ) //2901 is the maximum size for int when all values are 255
-            {
-                t_sum.create(src.cols, src.rows, CV_32SC1);
-                sum.create(h, w, CV_32SC1);
-            }
-            else
-            {
-                 //Use float to prevent overflow
-                t_sum.create(src.cols, src.rows, CV_32FC1);
-                sum.create(h, w, CV_32FC1);
-             }
-             t_sqsum.create(src.cols, src.rows, CV_32FC1);
-             sqsum.create(h, w, CV_32FC1);
-             depth = sum.depth();
-             int sum_offset = sum.offset / vlen;
-             int sqsum_offset = sqsum.offset / vlen;
-
-             vector<pair<size_t , const void *> > args;
-             args.push_back( make_pair( sizeof(cl_mem) , (void *)&src.data ));
-             args.push_back( make_pair( sizeof(cl_mem) , (void *)&t_sum.data ));
-             args.push_back( make_pair( sizeof(cl_mem) , (void *)&t_sqsum.data ));
-             args.push_back( make_pair( sizeof(cl_int) , (void *)&offset ));
-             args.push_back( make_pair( sizeof(cl_int) , (void *)&pre_invalid ));
-             args.push_back( make_pair( sizeof(cl_int) , (void *)&src.rows ));
-             args.push_back( make_pair( sizeof(cl_int) , (void *)&src.cols ));
-             args.push_back( make_pair( sizeof(cl_int) , (void *)&src.step ));
-             args.push_back( make_pair( sizeof(cl_int) , (void *)&t_sum.step));
-             size_t gt[3] = {((vcols + 1) / 2) * 256, 1, 1}, lt[3] = {256, 1, 1};
-             openCLExecuteKernel(src.clCxt, &imgproc_integral, "integral_cols", gt, lt, args, -1, depth);
-             args.clear();
-             args.push_back( make_pair( sizeof(cl_mem) , (void *)&t_sum.data ));
-             args.push_back( make_pair( sizeof(cl_mem) , (void *)&t_sqsum.data ));
-             args.push_back( make_pair( sizeof(cl_mem) , (void *)&sum.data ));
-             args.push_back( make_pair( sizeof(cl_mem) , (void *)&sqsum.data ));
-             args.push_back( make_pair( sizeof(cl_int) , (void *)&t_sum.rows ));
-             args.push_back( make_pair( sizeof(cl_int) , (void *)&t_sum.cols ));
-             args.push_back( make_pair( sizeof(cl_int) , (void *)&t_sum.step ));
-             args.push_back( make_pair( sizeof(cl_int) , (void *)&sum.step));
-             args.push_back( make_pair( sizeof(cl_int) , (void *)&sqsum.step));
-             args.push_back( make_pair( sizeof(cl_int) , (void *)&sum_offset));
-             args.push_back( make_pair( sizeof(cl_int) , (void *)&sqsum_offset));
-             size_t gt2[3] = {t_sum.cols  * 32, 1, 1}, lt2[3] = {256, 1, 1};
-             openCLExecuteKernel(src.clCxt, &imgproc_integral, "integral_rows", gt2, lt2, args, -1, depth);
+            int depth = src.depth() == CV_8U ? CV_32S : CV_64F;
+            int type = CV_MAKE_TYPE(depth, 1);
+
+            t_sum.create(src.cols, src.rows, type);
+            sum.create(h, w, type);
+
+            t_sqsum.create(src.cols, src.rows, CV_32FC1);
+            sqsum.create(h, w, CV_32FC1);
+
+            int sum_offset = sum.offset / vlen;
+            int sqsum_offset = sqsum.offset / vlen;
+
+            vector<pair<size_t , const void *> > args;
+            args.push_back( make_pair( sizeof(cl_mem) , (void *)&src.data ));
+            args.push_back( make_pair( sizeof(cl_mem) , (void *)&t_sum.data ));
+            args.push_back( make_pair( sizeof(cl_mem) , (void *)&t_sqsum.data ));
+            args.push_back( make_pair( sizeof(cl_int) , (void *)&offset ));
+            args.push_back( make_pair( sizeof(cl_int) , (void *)&pre_invalid ));
+            args.push_back( make_pair( sizeof(cl_int) , (void *)&src.rows ));
+            args.push_back( make_pair( sizeof(cl_int) , (void *)&src.cols ));
+            args.push_back( make_pair( sizeof(cl_int) , (void *)&src.step ));
+            args.push_back( make_pair( sizeof(cl_int) , (void *)&t_sum.step));
+            size_t gt[3] = {((vcols + 1) / 2) * 256, 1, 1}, lt[3] = {256, 1, 1};
+            openCLExecuteKernel(src.clCxt, &imgproc_integral, "integral_cols", gt, lt, args, -1, depth);
+
+            args.clear();
+            args.push_back( make_pair( sizeof(cl_mem) , (void *)&t_sum.data ));
+            args.push_back( make_pair( sizeof(cl_mem) , (void *)&t_sqsum.data ));
+            args.push_back( make_pair( sizeof(cl_mem) , (void *)&sum.data ));
+            args.push_back( make_pair( sizeof(cl_mem) , (void *)&sqsum.data ));
+            args.push_back( make_pair( sizeof(cl_int) , (void *)&t_sum.rows ));
+            args.push_back( make_pair( sizeof(cl_int) , (void *)&t_sum.cols ));
+            args.push_back( make_pair( sizeof(cl_int) , (void *)&t_sum.step ));
+            args.push_back( make_pair( sizeof(cl_int) , (void *)&sum.step));
+            args.push_back( make_pair( sizeof(cl_int) , (void *)&sqsum.step));
+            args.push_back( make_pair( sizeof(cl_int) , (void *)&sum_offset));
+            args.push_back( make_pair( sizeof(cl_int) , (void *)&sqsum_offset));
+            size_t gt2[3] = {t_sum.cols  * 32, 1, 1}, lt2[3] = {256, 1, 1};
+            openCLExecuteKernel(src.clCxt, &imgproc_integral, "integral_rows", gt2, lt2, args, -1, depth);
         }
 
         void integral(const oclMat &src, oclMat &sum)
@@ -1042,39 +1039,35 @@ namespace cv
 
             oclMat t_sum;
             int w = src.cols + 1, h = src.rows + 1;
-            int depth;
-            if(src.cols * src.rows <= 2901 * 2901)
-            {
-                t_sum.create(src.cols, src.rows, CV_32SC1);
-                sum.create(h, w, CV_32SC1);
-            }else
-            {
-                 t_sum.create(src.cols, src.rows, CV_32FC1);
-                 sum.create(h, w, CV_32FC1);
-             }
-             depth = sum.depth();
-             int sum_offset = sum.offset / vlen;
-             vector<pair<size_t , const void *> > args;
-             args.push_back( make_pair( sizeof(cl_mem) , (void *)&src.data ));
-             args.push_back( make_pair( sizeof(cl_mem) , (void *)&t_sum.data ));
-             args.push_back( make_pair( sizeof(cl_int) , (void *)&offset ));
-             args.push_back( make_pair( sizeof(cl_int) , (void *)&pre_invalid ));
-             args.push_back( make_pair( sizeof(cl_int) , (void *)&src.rows ));
-             args.push_back( make_pair( sizeof(cl_int) , (void *)&src.cols ));
-             args.push_back( make_pair( sizeof(cl_int) , (void *)&src.step ));
-             args.push_back( make_pair( sizeof(cl_int) , (void *)&t_sum.step));
-             size_t gt[3] = {((vcols + 1) / 2) * 256, 1, 1}, lt[3] = {256, 1, 1};
-             openCLExecuteKernel(src.clCxt, &imgproc_integral_sum, "integral_sum_cols", gt, lt, args, -1, depth);
-             args.clear();
-             args.push_back( make_pair( sizeof(cl_mem) , (void *)&t_sum.data ));
-             args.push_back( make_pair( sizeof(cl_mem) , (void *)&sum.data ));
-             args.push_back( make_pair( sizeof(cl_int) , (void *)&t_sum.rows ));
-             args.push_back( make_pair( sizeof(cl_int) , (void *)&t_sum.cols ));
-             args.push_back( make_pair( sizeof(cl_int) , (void *)&t_sum.step ));
-             args.push_back( make_pair( sizeof(cl_int) , (void *)&sum.step));
-             args.push_back( make_pair( sizeof(cl_int) , (void *)&sum_offset));
-             size_t gt2[3] = {t_sum.cols  * 32, 1, 1}, lt2[3] = {256, 1, 1};
-             openCLExecuteKernel(src.clCxt, &imgproc_integral_sum, "integral_sum_rows", gt2, lt2, args, -1, depth);
+            int depth = src.depth() == CV_8U ? CV_32S : CV_32F;
+            int type = CV_MAKE_TYPE(depth, 1);
+
+            t_sum.create(src.cols, src.rows, type);
+            sum.create(h, w, type);
+
+            int sum_offset = sum.offset / vlen;
+            vector<pair<size_t , const void *> > args;
+            args.push_back( make_pair( sizeof(cl_mem) , (void *)&src.data ));
+            args.push_back( make_pair( sizeof(cl_mem) , (void *)&t_sum.data ));
+            args.push_back( make_pair( sizeof(cl_int) , (void *)&offset ));
+            args.push_back( make_pair( sizeof(cl_int) , (void *)&pre_invalid ));
+            args.push_back( make_pair( sizeof(cl_int) , (void *)&src.rows ));
+            args.push_back( make_pair( sizeof(cl_int) , (void *)&src.cols ));
+            args.push_back( make_pair( sizeof(cl_int) , (void *)&src.step ));
+            args.push_back( make_pair( sizeof(cl_int) , (void *)&t_sum.step));
+            size_t gt[3] = {((vcols + 1) / 2) * 256, 1, 1}, lt[3] = {256, 1, 1};
+            openCLExecuteKernel(src.clCxt, &imgproc_integral_sum, "integral_sum_cols", gt, lt, args, -1, depth);
+
+            args.clear();
+            args.push_back( make_pair( sizeof(cl_mem) , (void *)&t_sum.data ));
+            args.push_back( make_pair( sizeof(cl_mem) , (void *)&sum.data ));
+            args.push_back( make_pair( sizeof(cl_int) , (void *)&t_sum.rows ));
+            args.push_back( make_pair( sizeof(cl_int) , (void *)&t_sum.cols ));
+            args.push_back( make_pair( sizeof(cl_int) , (void *)&t_sum.step ));
+            args.push_back( make_pair( sizeof(cl_int) , (void *)&sum.step));
+            args.push_back( make_pair( sizeof(cl_int) , (void *)&sum_offset));
+            size_t gt2[3] = {t_sum.cols  * 32, 1, 1}, lt2[3] = {256, 1, 1};
+            openCLExecuteKernel(src.clCxt, &imgproc_integral_sum, "integral_sum_rows", gt2, lt2, args, -1, depth);
         }
 
         /////////////////////// corner //////////////////////////////
diff --git a/modules/ocl/test/test_imgproc.cpp b/modules/ocl/test/test_imgproc.cpp
index 4d297a7a4f..86c750f371 100644
--- a/modules/ocl/test/test_imgproc.cpp
+++ b/modules/ocl/test/test_imgproc.cpp
@@ -579,7 +579,19 @@ TEST_P(cornerHarris, Mat)
 
 struct integral : ImgprocTestBase {};
 
-TEST_P(integral, Mat)
+TEST_P(integral, Mat1)
+{
+    for(int j = 0; j < LOOP_TIMES; j++)
+    {
+        random_roi();
+
+        cv::ocl::integral(clmat1_roi, cldst_roi);
+        cv::integral(mat1_roi, dst_roi);
+        Near(0);
+    }
+}
+
+TEST_P(integral, Mat2)
 {
     for(int j = 0; j < LOOP_TIMES; j++)
     {