| Age | Commit message (Collapse) | Author |
|---|
|
Bug: 313791883
Test: build
Change-Id: Ib4d58ebec02270f2c6dda4b307d03bd7e5526391
|
|
With this change we can now pass luma/chroma ptrs and stride information
to jpegencoderhelper class during compression. This by passes
intermediate copy whenever possible
Also updated fuzzer to incorporate 420 stride support
updated jpegr unit tests for more combinations of gamuts and unusual strides
Bug: 294218453
Test: ./ultrahdr_unit_test
Test: ./ultrahdr_enc_fuzzer
Change-Id: Ic50dd34b0c680618e73e0cb27f554b9bf8272e8f
|
|
- clang-format the code to a common guideline
- return status of areInputArgumentsValid call as-is
- rename variables for easier understanding
- reduce control code by assigning stride(s) and chroma ptr
after input validation
- if 420 color space is bt601, do not perform copy before
compressImage
TODO: ensure compressImage function takes luma ptr, chroma ptrs,
luma stride, chroma stride as well there by avoiding all
clobbering/intermediate copying
Bug:
Test: ./ultrahdr_unit_test
Change-Id: I9cda53419e1a7e8d68ddae571ea1bdd2ba5168f4
|
|
main
|
|
At high quality setting, less accurate integer method can have
larger psnr drops. Switch to default setting
Bug: 286617381
Test: ./libultrahdr_app -p inp_p010.yuv -y inp_420p.yuv -w 1920 -h 1080 -o 0
Change-Id: Ide126b87262e1c9a20edca7873b6ca27fc52b2cb
|
|
If calls to encode/decode failed, release the allocated memory
before returning the control to caller
Bug: 285546217
Test: ./ultrahdr_dec_fuzzer
Test: ./ultrahdr_enc_fuzzer
Change-Id: I276c31cc56656aa41845a16f5d28783bc3adc772
|
|
jpeg_write_raw_data() processes one MCU row per call, and thus one must
pass buffer of at least max_v_samp_factor * DCTSIZE scanlines. The
buffer must be large enough to hold the actual data plus padding
to DCT-block boundaries.
Bug: 284117683
Test: ./ultrahdr_enc_fuzzer
Change-Id: I993773817bf3805463bb21bb977624d6c2d45a0b
|
|
JPEG/R library uses jpeg-turbo for JPEG encoding, which runs DCT transform on block size of 16x16 for luma, and 8x8 for chroma. The resolution in the bug report is not 16-aligned and it results in null pointer dereference for the last line in jpeg-turbo.
The original checking for 8-alignment width was wrong (should check
16-alignment). jpeg-turbo has some edge case handling for this case, and
it requires some extra room at the end of input. This change removed the checking criteria by adding a
padding zero method. A reason size of the padding zeros is a CB block,
which is 8x8, 64 bytes.
Bug: 277982036
Test: CTS: ImageReaderTest#testJpegR, uint test: jpegrencoderhelper_test.cpp
Change-Id: I1313a002db6d4bc63b32dc3dd3d6ccdf06779149
|
|
Test: build
Bug: b/264715926
Change-Id: I227fb5960f8fc7e13aae354bf77ec033850faf10
|