Add support for multiple regions of interest (ROI) for batch JPEG decoding (#126)

AryanSalmanpour · web-flow · commit 63637ce8f6df · 2025-03-05T17:06:04.000-05:00
[ROCm/rocjpeg commit: 692d4bb]
diff --git a/projects/rocjpeg/CHANGELOG.md b/projects/rocjpeg/CHANGELOG.md
@@ -3,10 +3,11 @@
 Documentation for rocJPEG is available at
 [https://rocm.docs.amd.com/projects/rocJPEG/en/latest/](https://rocm.docs.amd.com/projects/rocJPEG/en/latest/)
 
-## (Unreleased) rocJPEG 0.9.0
+## (Unreleased) rocJPEG 0.10.0
 
 ## Changed
 * Readme - cleanup and updates to pre-reqs
+* The `decode_params` argument of the `rocJpegDecodeBatched` API is now an array of `RocJpegDecodeParams` structs representing the decode parameters for the batch of JPEG images.
 
 ## Removed
 * Dev Package - No longer installs pkg-config
diff --git a/projects/rocjpeg/CMakeLists.txt b/projects/rocjpeg/CMakeLists.txt
@@ -40,7 +40,7 @@ set(CMAKE_CXX_STANDARD 17)
 set(CMAKE_CXX_STANDARD_REQUIRED On)
 
 # NOTE: Match version with api/rocjpeg_version.h
-set(VERSION "0.9.0")
+set(VERSION "0.10.0")
 # Set Project Version and Language
 project(rocjpeg VERSION ${VERSION} LANGUAGES CXX)
 
diff --git a/projects/rocjpeg/api/rocjpeg.h b/projects/rocjpeg/api/rocjpeg.h
@@ -319,12 +319,12 @@ RocJpegStatus ROCJPEGAPI rocJpegDecode(RocJpegHandle handle, RocJpegStreamHandle
  * @ingroup group_amd_rocjpeg
  * @brief Decodes a batch of JPEG images using the rocJPEG library.
  *
- * Decodes a batch of JPEG images using the rocJPEG library.
+ * This function decodes a batch of JPEG images using the rocJPEG library.
  *
  * @param handle The rocJPEG handle.
  * @param jpeg_stream_handles An array of rocJPEG stream handles representing the input JPEG streams.
  * @param batch_size The number of JPEG streams in the batch.
- * @param decode_params The decode parameters for the JPEG decoding process.
+ * @param decode_params An array of RocJpegDecodeParams structs representing the decode parameters for the batch of JPEG images.
  * @param destinations An array of rocJPEG images representing the output decoded images.
  * @return The status of the JPEG decoding operation.
  */
diff --git a/projects/rocjpeg/api/rocjpeg_version.h b/projects/rocjpeg/api/rocjpeg_version.h
@@ -34,7 +34,7 @@ THE SOFTWARE.
 extern "C" {
 #endif
 #define ROCJPEG_MAJOR_VERSION 0
-#define ROCJPEG_MINOR_VERSION 9
+#define ROCJPEG_MINOR_VERSION 10
 #define ROCJPEG_MICRO_VERSION 0
 
 
diff --git a/projects/rocjpeg/samples/jpegDecodeBatched/jpegdecodebatched.cpp b/projects/rocjpeg/samples/jpegDecodeBatched/jpegdecodebatched.cpp
@@ -49,6 +49,7 @@ int main(int argc, char **argv) {
     RocJpegImage output_image = {};
     std::vector<RocJpegImage> output_images;
     RocJpegDecodeParams decode_params = {};
+    std::vector<RocJpegDecodeParams> decode_params_batch;
     RocJpegUtils rocjpeg_utils;
     std::vector<std::string> base_file_names;
     std::vector<RocJpegStreamHandle> rocjpeg_stream_handles_for_current_batch;
@@ -61,15 +62,11 @@ int main(int argc, char **argv) {
     uint64_t num_jpegs_with_unknown_subsampling = 0;
     uint64_t num_jpegs_with_unsupported_resolution = 0;
     int current_batch_size = 0;
-
-    RocJpegUtils::ParseCommandLine(input_path, output_file_path, save_images, device_id, rocjpeg_backend, decode_params, nullptr, &batch_size, argc, argv);
-    
     bool is_roi_valid = false;
     uint32_t roi_width;
     uint32_t roi_height;
-    roi_width = decode_params.crop_rectangle.right - decode_params.crop_rectangle.left;
-    roi_height = decode_params.crop_rectangle.bottom - decode_params.crop_rectangle.top;
-    
+
+    RocJpegUtils::ParseCommandLine(input_path, output_file_path, save_images, device_id, rocjpeg_backend, decode_params, nullptr, &batch_size, argc, argv);
     if (!RocJpegUtils::GetFilePaths(input_path, file_paths, is_dir, is_file)) {
         std::cerr << "ERROR: Failed to get input file paths!" << std::endl;
         return EXIT_FAILURE;
@@ -89,6 +86,7 @@ int main(int argc, char **argv) {
 
     batch_images.resize(batch_size);
     output_images.resize(batch_size);
+    decode_params_batch.resize(batch_size, decode_params);
     prior_channel_sizes.resize(batch_size, std::vector<uint32_t>(ROCJPEG_MAX_COMPONENT, 0));
     widths.resize(batch_size, std::vector<uint32_t>(ROCJPEG_MAX_COMPONENT, 0));
     heights.resize(batch_size, std::vector<uint32_t>(ROCJPEG_MAX_COMPONENT, 0));
@@ -134,10 +132,6 @@ int main(int argc, char **argv) {
 
             CHECK_ROCJPEG(rocJpegGetImageInfo(rocjpeg_handle, rocjpeg_stream_handles[index], &num_components, &temp_subsampling, temp_widths.data(), temp_heights.data()));
 
-            if (roi_width > 0 && roi_height > 0 && roi_width <= widths[index][0] && roi_height <= heights[index][0]) {
-                is_roi_valid = true; 
-            }
-
             rocjpeg_utils.GetChromaSubsamplingStr(temp_subsampling, chroma_sub_sampling);
             if (temp_widths[0] < 64 || temp_heights[0] < 64) {
                 if (is_dir) {
@@ -164,7 +158,7 @@ int main(int argc, char **argv) {
                 }
             }
 
-            if (rocjpeg_utils.GetChannelPitchAndSizes(decode_params, temp_subsampling, temp_widths.data(), temp_heights.data(), num_channels, output_images[current_batch_size], channel_sizes)) {
+            if (rocjpeg_utils.GetChannelPitchAndSizes(decode_params_batch[index], temp_subsampling, temp_widths.data(), temp_heights.data(), num_channels, output_images[current_batch_size], channel_sizes)) {
                 std::cerr << "ERROR: Failed to get the channel pitch and sizes" << std::endl;
                 return EXIT_FAILURE;
             }
@@ -191,7 +185,7 @@ int main(int argc, char **argv) {
         double time_per_batch_in_milli_sec = 0;
         if (current_batch_size > 0) {
             auto start_time = std::chrono::high_resolution_clock::now();
-            CHECK_ROCJPEG(rocJpegDecodeBatched(rocjpeg_handle, rocjpeg_stream_handles_for_current_batch.data(), current_batch_size, &decode_params, output_images.data()));
+            CHECK_ROCJPEG(rocJpegDecodeBatched(rocjpeg_handle, rocjpeg_stream_handles_for_current_batch.data(), current_batch_size, decode_params_batch.data(), output_images.data()));
             auto end_time = std::chrono::high_resolution_clock::now();
             time_per_batch_in_milli_sec = std::chrono::duration<double, std::milli>(end_time - start_time).count();
         }
@@ -207,12 +201,15 @@ int main(int argc, char **argv) {
             for (int b = 0; b < current_batch_size; b++) {
                 std::string image_save_path = output_file_path;
                 //if ROI is present, need to pass roi_width and roi_height
+                roi_width = decode_params_batch[b].crop_rectangle.right - decode_params_batch[b].crop_rectangle.left;
+                roi_height = decode_params_batch[b].crop_rectangle.bottom - decode_params_batch[b].crop_rectangle.top;
+                is_roi_valid = (roi_width > 0 && roi_height > 0 && roi_width <= widths[b][0] && roi_height <= heights[b][0]) ? true : false;
                 uint32_t width = is_roi_valid ? roi_width : widths[b][0];
                 uint32_t height = is_roi_valid ? roi_height : heights[b][0];
                 if (is_dir) {
-                    rocjpeg_utils.GetOutputFileExt(decode_params.output_format, base_file_names[b], width, height, subsamplings[b], image_save_path);
+                    rocjpeg_utils.GetOutputFileExt(decode_params_batch[b].output_format, base_file_names[b], width, height, subsamplings[b], image_save_path);
                 }
-                rocjpeg_utils.SaveImage(image_save_path, &output_images[b], width, height, subsamplings[b], decode_params.output_format);
+                rocjpeg_utils.SaveImage(image_save_path, &output_images[b], width, height, subsamplings[b], decode_params_batch[b].output_format);
             }
         }
 
diff --git a/projects/rocjpeg/samples/jpegDecodePerf/jpegdecodeperf.cpp b/projects/rocjpeg/samples/jpegDecodePerf/jpegdecodeperf.cpp
@@ -50,8 +50,6 @@ void DecodeImages(DecodeInfo &decode_info, RocJpegUtils rocjpeg_utils, RocJpegDe
     bool is_roi_valid = false;
     uint32_t roi_width;
     uint32_t roi_height;
-    roi_width = decode_params.crop_rectangle.right - decode_params.crop_rectangle.left;
-    roi_height = decode_params.crop_rectangle.bottom - decode_params.crop_rectangle.top;
     uint8_t num_components;
     uint32_t channel_sizes[ROCJPEG_MAX_COMPONENT] = {};
     std::string chroma_sub_sampling = "";
@@ -65,6 +63,7 @@ void DecodeImages(DecodeInfo &decode_info, RocJpegUtils rocjpeg_utils, RocJpegDe
     std::vector<std::vector<uint32_t>> prior_channel_sizes(batch_size, std::vector<uint32_t>(ROCJPEG_MAX_COMPONENT, 0));
     std::vector<RocJpegChromaSubsampling> subsamplings(batch_size);
     std::vector<RocJpegImage> output_images(batch_size);
+    std::vector<RocJpegDecodeParams> decode_params_batch(batch_size, decode_params);
     std::vector<std::string> base_file_names(batch_size);
     std::vector<RocJpegStreamHandle> rocjpeg_stream_handles(batch_size);
     std::vector<uint32_t> temp_widths(ROCJPEG_MAX_COMPONENT, 0);
@@ -105,9 +104,6 @@ void DecodeImages(DecodeInfo &decode_info, RocJpegUtils rocjpeg_utils, RocJpegDe
             }
 
             CHECK_ROCJPEG(rocJpegGetImageInfo(decode_info.rocjpeg_handle, decode_info.rocjpeg_stream_handles[index], &num_components, &temp_subsampling, temp_widths.data(), temp_heights.data()));
-            if (roi_width > 0 && roi_height > 0 && roi_width <= temp_widths[0] && roi_height <= temp_heights[0]) {
-                is_roi_valid = true; 
-            }
 
             rocjpeg_utils.GetChromaSubsamplingStr(temp_subsampling, chroma_sub_sampling);
             if (temp_widths[0] < 64 || temp_heights[0] < 64) {
@@ -125,7 +121,7 @@ void DecodeImages(DecodeInfo &decode_info, RocJpegUtils rocjpeg_utils, RocJpegDe
                 continue;
             }
 
-            if (rocjpeg_utils.GetChannelPitchAndSizes(decode_params, temp_subsampling, temp_widths.data(), temp_heights.data(), num_channels, output_images[current_batch_size], channel_sizes)) {
+            if (rocjpeg_utils.GetChannelPitchAndSizes(decode_params_batch[index], temp_subsampling, temp_widths.data(), temp_heights.data(), num_channels, output_images[current_batch_size], channel_sizes)) {
                 std::cerr << "ERROR: Failed to get the channel pitch and sizes" << std::endl;
                 return;
             }
@@ -153,7 +149,7 @@ void DecodeImages(DecodeInfo &decode_info, RocJpegUtils rocjpeg_utils, RocJpegDe
         double time_per_batch_in_milli_sec = 0;
         if (current_batch_size > 0) {
             auto start_time = std::chrono::high_resolution_clock::now();
-            CHECK_ROCJPEG(rocJpegDecodeBatched(decode_info.rocjpeg_handle, rocjpeg_stream_handles.data(), current_batch_size, &decode_params, output_images.data()));
+            CHECK_ROCJPEG(rocJpegDecodeBatched(decode_info.rocjpeg_handle, rocjpeg_stream_handles.data(), current_batch_size, decode_params_batch.data(), output_images.data()));
             auto end_time = std::chrono::high_resolution_clock::now();
             time_per_batch_in_milli_sec = std::chrono::duration<double, std::milli>(end_time - start_time).count();
         }
@@ -169,6 +165,9 @@ void DecodeImages(DecodeInfo &decode_info, RocJpegUtils rocjpeg_utils, RocJpegDe
             for (int b = 0; b < current_batch_size; b++) {
                 std::string image_save_path = output_file_path;
                 //if ROI is present, need to pass roi_width and roi_height
+                roi_width = decode_params_batch[b].crop_rectangle.right - decode_params_batch[b].crop_rectangle.left;
+                roi_height = decode_params_batch[b].crop_rectangle.bottom - decode_params_batch[b].crop_rectangle.top;
+                is_roi_valid = (roi_width > 0 && roi_height > 0 && roi_width <= widths[b][0] && roi_height <= heights[b][0]) ? true : false;
                 uint32_t width = is_roi_valid ? roi_width : widths[b][0];
                 uint32_t height = is_roi_valid ? roi_height : heights[b][0];
                 rocjpeg_utils.GetOutputFileExt(decode_params.output_format, base_file_names[b], width, height, subsamplings[b], image_save_path);
diff --git a/projects/rocjpeg/src/rocjpeg_decoder.cpp b/projects/rocjpeg/src/rocjpeg_decoder.cpp
@@ -216,7 +216,7 @@ RocJpegStatus RocJpegDecoder::DecodeBatched(RocJpegStreamHandle *jpeg_streams, i
             jpeg_streams_params[j] = std::move(*jpeg_stream_params);
         }
 
-        CHECK_ROCJPEG(jpeg_vaapi_decoder_.SubmitDecodeBatched(jpeg_streams_params.data() + i, current_batch_size, decode_params, current_surface_ids.data() + i));
+        CHECK_ROCJPEG(jpeg_vaapi_decoder_.SubmitDecodeBatched(jpeg_streams_params.data() + i, current_batch_size, &decode_params[i], current_surface_ids.data() + i));
 
         for (int k = 0; k < current_batch_size; k++) {
             HipInteropDeviceMem hip_interop_dev_mem = {};
@@ -231,8 +231,8 @@ RocJpegStatus RocJpegDecoder::DecodeBatched(RocJpegStreamHandle *jpeg_streams, i
             bool is_roi_valid = false;
             uint32_t roi_width;
             uint32_t roi_height;
-            roi_width = decode_params->crop_rectangle.right - decode_params->crop_rectangle.left;
-            roi_height = decode_params->crop_rectangle.bottom - decode_params->crop_rectangle.top;
+            roi_width = decode_params[k + i].crop_rectangle.right - decode_params[k + i].crop_rectangle.left;
+            roi_height = decode_params[k + i].crop_rectangle.bottom - decode_params[k + i].crop_rectangle.top;
     
             if (roi_width > 0 && roi_height > 0 && roi_width <= jpeg_stream_params->picture_parameter_buffer.picture_width && roi_height <= jpeg_stream_params->picture_parameter_buffer.picture_height) {
                 is_roi_valid = true;
@@ -247,38 +247,38 @@ RocJpegStatus RocJpegDecoder::DecodeBatched(RocJpegStreamHandle *jpeg_streams, i
                 is_roi_valid = false;
             }
 
-            switch (decode_params->output_format) {
+            switch (decode_params[k + i].output_format) {
                 case ROCJPEG_OUTPUT_NATIVE:
                     // Copy the native decoded output buffers from interop memory directly to the destination buffers
                     CHECK_ROCJPEG(GetChromaHeight(hip_interop_dev_mem.surface_format, picture_height, chroma_height));
                     // Copy Luma (first channel) for any surface format
-                    CHECK_ROCJPEG(CopyChannel(hip_interop_dev_mem, picture_height, 0, &destinations[k + i], decode_params, is_roi_valid));
+                    CHECK_ROCJPEG(CopyChannel(hip_interop_dev_mem, picture_height, 0, &destinations[k + i], &decode_params[k + i], is_roi_valid));
                     if (hip_interop_dev_mem.surface_format == VA_FOURCC_NV12) {
                         // Copy the second channel (UV interleaved) for NV12
-                        CHECK_ROCJPEG(CopyChannel(hip_interop_dev_mem, chroma_height, 1, &destinations[k + i], decode_params, is_roi_valid));
+                        CHECK_ROCJPEG(CopyChannel(hip_interop_dev_mem, chroma_height, 1, &destinations[k + i], &decode_params[k + i], is_roi_valid));
                     } else if (hip_interop_dev_mem.surface_format == VA_FOURCC_444P ||
                             hip_interop_dev_mem.surface_format == VA_FOURCC_422V) {
                         // Copy the second and third channels for YUV444 and YUV440 (i.e., YUV422V)
-                        CHECK_ROCJPEG(CopyChannel(hip_interop_dev_mem, chroma_height, 1, &destinations[k + i], decode_params, is_roi_valid));
-                        CHECK_ROCJPEG(CopyChannel(hip_interop_dev_mem, chroma_height, 2, &destinations[k + i], decode_params, is_roi_valid));
+                        CHECK_ROCJPEG(CopyChannel(hip_interop_dev_mem, chroma_height, 1, &destinations[k + i], &decode_params[k + i], is_roi_valid));
+                        CHECK_ROCJPEG(CopyChannel(hip_interop_dev_mem, chroma_height, 2, &destinations[k + i], &decode_params[k + i], is_roi_valid));
                     }
                     break;
                 case ROCJPEG_OUTPUT_YUV_PLANAR:
                     CHECK_ROCJPEG(GetChromaHeight(hip_interop_dev_mem.surface_format, picture_height, chroma_height));
                     CHECK_ROCJPEG(GetPlanarYUVOutputFormat(hip_interop_dev_mem, picture_width,
-                                                        picture_height, chroma_height, &destinations[k + i], decode_params, is_roi_valid));
+                                                        picture_height, chroma_height, &destinations[k + i], &decode_params[k + i], is_roi_valid));
                     break;
                 case ROCJPEG_OUTPUT_Y:
                     CHECK_ROCJPEG(GetYOutputFormat(hip_interop_dev_mem, picture_width,
-                                                picture_height, &destinations[k + i], decode_params, is_roi_valid));
+                                                picture_height, &destinations[k + i], &decode_params[k + i], is_roi_valid));
                     break;
                 case ROCJPEG_OUTPUT_RGB:
                     CHECK_ROCJPEG(ColorConvertToRGB(hip_interop_dev_mem, picture_width,
-                                                            picture_height, &destinations[k + i], decode_params, is_roi_valid));
+                                                            picture_height, &destinations[k + i], &decode_params[k + i], is_roi_valid));
                     break;
                 case ROCJPEG_OUTPUT_RGB_PLANAR:
                     CHECK_ROCJPEG(ColorConvertToRGBPlanar(hip_interop_dev_mem, picture_width,
-                                                            picture_height, &destinations[k + i], decode_params, is_roi_valid));
+                                                            picture_height, &destinations[k + i], &decode_params[k + i], is_roi_valid));
                     break;
                 default:
                     break;
diff --git a/projects/rocjpeg/src/rocjpeg_vaapi_decoder.cpp b/projects/rocjpeg/src/rocjpeg_vaapi_decoder.cpp
