| January 7, 2010 11:00 PM PST | |
DPD200130124 - JPEG color conversion function request
Added support for RGBA as an input format for the IPP JPEG encoder and RGBA/BGRA as an output format for the IPP JPEG decoder. This functionality is also available in the UIC JPEG codec. This function helps avoid a memory copy from RGBA to RGB format.
DPD200134755 - incorrect decoding for IPP MJPEG sample in UMC
Fixed an issue where the last (bottom right) 8x8 fragment of an image might be incorrectly decoded and instead there is a solid gray 8x8 area.
DPD200134712 - distored JPEG output
Opening a BMP file and saving as JPEG using the settings: Bas, 444, Huffman, RGB, quality=75 results in a grayscale and distorted JPEG file.
DPD200134360 - ippiDecodeHuffman8x8_Direct_JPEG_1u163_C1 read beyong input data
An issue with the above function when decoding a JPEG file under certain conditions; the function reads beyond the input data. This is reported by Valgrind or when using the Electric Fence tool. This has been tested on Linux both 32-bit and 64-bit. The "unoptimized" variants (i.e. 'px' and 'mx') do not exhibit this problem. Although not tested, the problem may also occur on Windows.
DPD200134372 - read error in 'ippiDCTQuantInv8x8To2x2LS_JPEG_16s8u_C1R' for JPEG decoding
A problem in the above function exists when reducing JPEG images to a lower resolution. The problem occurs when 'ippiDCTQuantInv8x8To2x2LS_JPEG_16s8u_C1R' and an inverse DCT, inverse quantization and level-shifting is focused on the upper-left 2x2 part of an 8x8 block. The function reads "too far" in the *destination* buffer. This can cause a read access violation. Observed on *both* Linux and Windows and both 32-bit and 64-bit optimizations. The "unoptimized" 'px' (32-bit) and 'mx' (64-bit) variants do not exhibit the problem.
DPD200134376 - bug in colors for planar destinations at lower resolution option in 'JPEGView'
Not a bug but a request for a new mode: decoding of reduced resolution for planar output image format. When choosing to decode to smaller resolutions and extract native data in planar YCbCr format, the color planes seem to be wrongly handled as the resulting image shows color errors. Only the left half of image contains color, that is incorrect.
DPD200133366 - resolution not saved in JPEG
The sample code encoder/decoder does not pass Image Resolution (DPI) for JPEG.
This article applies to: Intel® Integrated Performance Primitives Knowledge Base
For more complete information about compiler optimizations, see our Optimization Notice.
Comments (0) 
Trackbacks (0)
Leave a comment 
Paul Fischer (Intel)
|

