From 82e2792ab5cfaafc7d0c2ca7c22cecd56dba1c67 Mon Sep 17 00:00:00 2001
From: cudawarped <12133430+cudawarped@users.noreply.github.com>
Date: Mon, 6 Mar 2023 12:19:05 +0200
Subject: [PATCH] cudacodec: ensure output frame resolution is display not
 coded size

---
 modules/cudacodec/src/video_parser.cpp |  8 +++---
 modules/cudacodec/test/test_video.cpp  | 37 ++++++++++++++++++++++++++
 2 files changed, 41 insertions(+), 4 deletions(-)

diff --git a/modules/cudacodec/src/video_parser.cpp b/modules/cudacodec/src/video_parser.cpp
index 8bccd065a8d..e9ce56339bc 100644
--- a/modules/cudacodec/src/video_parser.cpp
+++ b/modules/cudacodec/src/video_parser.cpp
@@ -122,16 +122,16 @@ int CUDAAPI cv::cudacodec::detail::VideoParser::HandleVideoSequence(void* userDa
         newFormat.ulHeight = format->coded_height;
         newFormat.fps = format->frame_rate.numerator / static_cast<float>(format->frame_rate.denominator);
         newFormat.targetSz = thiz->videoDecoder_->getTargetSz();
-        newFormat.width = newFormat.targetSz.width ? newFormat.targetSz.width : format->coded_width;
-        newFormat.height = newFormat.targetSz.height ? newFormat.targetSz.height : format->coded_height;
         newFormat.srcRoi = thiz->videoDecoder_->getSrcRoi();
         if (newFormat.srcRoi.empty()) {
-            format->display_area.right = format->coded_width;
-            format->display_area.bottom = format->coded_height;
             newFormat.displayArea = Rect(Point(format->display_area.left, format->display_area.top), Point(format->display_area.right, format->display_area.bottom));
+            if (newFormat.targetSz.empty())
+                newFormat.targetSz = Size((format->display_area.right - format->display_area.left), (format->display_area.bottom - format->display_area.top));
         }
         else
             newFormat.displayArea = newFormat.srcRoi;
+        newFormat.width = newFormat.targetSz.width ? newFormat.targetSz.width : format->coded_width;
+        newFormat.height = newFormat.targetSz.height ? newFormat.targetSz.height : format->coded_height;
         newFormat.targetRoi = thiz->videoDecoder_->getTargetRoi();
         newFormat.ulNumDecodeSurfaces = min(!thiz->allowFrameDrop_ ? max(thiz->videoDecoder_->nDecodeSurfaces(), static_cast<int>(format->min_num_decode_surfaces)) :
             format->min_num_decode_surfaces * 2, 32);
diff --git a/modules/cudacodec/test/test_video.cpp b/modules/cudacodec/test/test_video.cpp
index b9b4e9f25c6..52105536225 100644
--- a/modules/cudacodec/test/test_video.cpp
+++ b/modules/cudacodec/test/test_video.cpp
@@ -58,6 +58,10 @@ PARAM_TEST_CASE(Scaling, cv::cuda::DeviceInfo, std::string, Size2f, Rect2f, Rect
 {
 };
 
+struct DisplayResolution : testing::TestWithParam<cv::cuda::DeviceInfo>
+{
+};
+
 PARAM_TEST_CASE(Video, cv::cuda::DeviceInfo, std::string)
 {
 };
@@ -223,6 +227,37 @@ CUDA_TEST_P(Scaling, Reader)
     ASSERT_LT(mae, 2.35);
 }
 
+CUDA_TEST_P(DisplayResolution, Reader)
+{
+    cv::cuda::setDevice(GetParam().deviceID());
+    std::string inputFile = std::string(cvtest::TS::ptr()->get_data_path()) + "../cv/video/1920x1080.avi";
+    const Rect displayArea(0, 0, 1920, 1080);
+    GpuMat frame;
+
+    {
+        // verify the output frame is the diplay size (1920x1080) and not the coded size (1920x1088)
+        cv::Ptr<cv::cudacodec::VideoReader> reader = cv::cudacodec::createVideoReader(inputFile);
+        reader->set(cudacodec::ColorFormat::GRAY);
+        ASSERT_TRUE(reader->nextFrame(frame));
+        const cudacodec::FormatInfo format = reader->format();
+        ASSERT_TRUE(format.displayArea == displayArea);
+        ASSERT_TRUE(frame.size() == displayArea.size() && frame.size() == format.targetSz);
+    }
+
+    {
+        // extra check to verify display frame has not been post-processed and is just a cropped version of the coded sized frame
+        cudacodec::VideoReaderInitParams params;
+        params.srcRoi = Rect(0, 0, 1920, 1088);
+        cv::Ptr<cv::cudacodec::VideoReader> readerCodedSz = cv::cudacodec::createVideoReader(inputFile, {}, params);
+        readerCodedSz->set(cudacodec::ColorFormat::GRAY);
+        GpuMat frameCodedSz;
+        ASSERT_TRUE(readerCodedSz->nextFrame(frameCodedSz));
+        const cudacodec::FormatInfo formatCodedSz = readerCodedSz->format();
+        const double err = cv::cuda::norm(frame, frameCodedSz(displayArea), NORM_INF);
+        ASSERT_TRUE(err == 0);
+    }
+}
+
 CUDA_TEST_P(Video, Reader)
 {
     cv::cuda::setDevice(GET_PARAM(0).deviceID());
@@ -664,6 +699,8 @@ INSTANTIATE_TEST_CASE_P(CUDA_Codec, CheckSet, testing::Combine(
 INSTANTIATE_TEST_CASE_P(CUDA_Codec, Scaling, testing::Combine(
     ALL_DEVICES, testing::Values(VIDEO_SRC_SCALING), testing::Values(TARGET_SZ), testing::Values(SRC_ROI), testing::Values(TARGET_ROI)));
 
+INSTANTIATE_TEST_CASE_P(CUDA_Codec, DisplayResolution, ALL_DEVICES);
+
 #define VIDEO_SRC_R "highgui/video/big_buck_bunny.mp4", "cv/video/768x576.avi", "cv/video/1920x1080.avi", "highgui/video/big_buck_bunny.avi", \
     "highgui/video/big_buck_bunny.h264", "highgui/video/big_buck_bunny.h265", "highgui/video/big_buck_bunny.mpg", \
     "highgui/video/sample_322x242_15frames.yuv420p.libvpx-vp9.mp4", "highgui/video/sample_322x242_15frames.yuv420p.libaom-av1.mp4", \