The files fts_18_1805.pdf and fts_18_1812.pdf are rendered by Ghostscript 8.71 and head (r11532) with the images incorrect when compared to Adobe Acrobat 9.3.3. With 8.71 all four images are wrong, with head two of the four are correct but two are still wrong, see the attached screenshot.
This bug is caused by an image with opacity layer that cannot be processed correctly by Jasper. No hacking of the output stream Can restore the image. This includes experiments with DeviceN and tint transform functions.
The pdf interpreter needs to detect SMaskInData true in the file. In this case, hastransparency should be true and the pdf14 device pushed. The image should be handled as a 3x type and the alpha channel will be interleaved with the image data. Jasper processes the file just fine, it is GS that is not using the alpha channel properly nor paying attention to the SMaskInData setting.
The following files are also affected by this bug. fts_17_1700_a4.pdf fts_17_1701_a4.pdf
Created attachment 6666 [details] Partial patch Skip opacity and unknown components during reading of image data. This improves the image quality (compared to the previous state that interpreted an opacity channel as yet another data sample) and serves as a transition to full support of opacity.
The patch has been committed as a rev. 11637.
Created attachment 6679 [details] experimental patch This is rather inefficient implementation of opacity in the image that uses separate streams for the image data and opacity data. It also buffers opacity data in an reusable stream. This patch is attached in hope that some parts of the code may be usable. Proper implementation should use GS extensions that permit interleaved image and opacity data.
*** Bug 691516 has been marked as a duplicate of this bug. ***
Created attachment 6824 [details] patch Implement /SMaskInData using images with /SMask and separate streams for the mask and image data. This patch is essentially the same as the "experimental patch" posted earlier with a few bug fixes. The interface to /JPXDecode filter has changed. It takes now /Alpha parameter, which controls whether opacity or image data stream is extracted. Our /ImageType 103 seems to have issues and my attempts to use it directly have failed. The patch has been committed as a rev. 11845. The commit introduces a PS error on pdfwrite device in fts_18_1805.pdf and a few other files with the same image. However, this image has never been combined with SMask before and the actual bug is likely to predate the commit.
Current version is working wit both Jasper and Luratech libraries. Remaining issues are tracked as separate bugs.