Decode one row of tiles at a time for image that has tiles
Add an api to IMediaMetadataRetriever to decode image rect.
It will reuse the same full frame IMemory, and decode only
the requested rect. For now, StagefrightMetadataRetriever
will only allow decoding of rect that's a full row of tiles,
and the requested must be issued sequentially (i.e. no
arbitrary rects). When the extract side is fixed to allow
seeking by tiles, it can be extended to allow arbitrary
rects.
This allows HeifDecoderImpl (on client side) to start
processing the getScanlines in parallel with the decoding.
Test: CTS MediaMetadataRetrieverTest;
Manual testing of HEIF decoding of files with or without tiles;
Manual testing of HEIF thumbnails generation in Downloads app.
bug: 78475896
Change-Id: I820b21cdf33f80593ee6092d8e1ba68b3beb65dd
diff --git a/media/libmedia/mediametadataretriever.cpp b/media/libmedia/mediametadataretriever.cpp
index c10a907..e61b04d 100644
--- a/media/libmedia/mediametadataretriever.cpp
+++ b/media/libmedia/mediametadataretriever.cpp
@@ -166,6 +166,19 @@
return mRetriever->getImageAtIndex(index, colorFormat, metaOnly, thumbnail);
}
+sp<IMemory> MediaMetadataRetriever::getImageRectAtIndex(
+ int index, int colorFormat, int left, int top, int right, int bottom) {
+ ALOGV("getImageRectAtIndex: index(%d) colorFormat(%d) rect {%d, %d, %d, %d}",
+ index, colorFormat, left, top, right, bottom);
+ Mutex::Autolock _l(mLock);
+ if (mRetriever == 0) {
+ ALOGE("retriever is not initialized");
+ return NULL;
+ }
+ return mRetriever->getImageRectAtIndex(
+ index, colorFormat, left, top, right, bottom);
+}
+
status_t MediaMetadataRetriever::getFrameAtIndex(
std::vector<sp<IMemory> > *frames,
int frameIndex, int numFrames, int colorFormat, bool metaOnly) {