DM368 mjpeg decoder with XDM_YUV_420SP output

Hi Huo,

Are you using DVSDK? If yes, may I know which version you are using?

Regards, RSV

Hi RSV,

Thansk for you reply!

I'm using dvsdk_2_10_00_17, and had updated newest codecs from TI website.

Regards, Huo

Hi Chengping,

I am having the same problem with forceChromaFormat set to XDM_YUV_420SP when decoding a 1920x1080 jpeg file.  It appears the small random dark blocks are on the Y planar only.  I am wondering if you have solved the problem.

The problem still exists even I tried decoding the 1920x1080 JPEG file captured using XDM's jpegenc (01_00_00_09) with inputChromaFormat set to XDM_YUV_420SP.

Regards,

Jimmy

Finally found a workaround to get those random dark blocks (see attached picture) removed from the luma planar of the jpeg decoder output.  It turns out that I had to turn off V4L2 capture (by issuing VIDIOC_STREAMOFF ioctl call) and waited for a few hundred milliseconds before starting the JPEG decoder.  I don't know how V4L2 capture and JPEG decoder got related to each other to cause conflicts.  Maybe TI engineers can take a look into it.

Jimmy

I have not fixed this issue. I guess this is a bug in TI mjpeg decoder.

I have the same problem with forceChromaFormat set to XDM_YUV_420SP when decoding a 1280x720 jpeg file(DM368)

How to resolve this problem?

Can you please provide input JPEG file?

Regards, RSV

I think All jpeg files have the same problem(random small block).

We are using captured jpeg file(720P) from HDMI input.

This problem appears Jpeg decoder forceChromaFormat value only set to XDM_YUV_420SP(forceChromaFormat).

Attached file is sample file(1280x720). If do not appear this problem, plz test repeatly.

hi, all

i want to ask if you got this problem solved.

i got the same problem as you did.

regards, Mike

Thanks.

the jpegdec codec versioin is just the newest one, say 1.00.00.10 as you can see here http://software-dl.ti.com/dsps/dsps_public_sw/codecs/DM36x/index_FDS.html

i have done 800*480 jpegdec and i believe the input buff is the same as the file we can see in Windows.

it's random thing when i have don 100 times jpegdec and just found the black block about 10 times.

it's very annoy and need your help.

regards, Mike

thanks in advance

Hi,

We are trying to reproduce the issue at our end using standalone testapp provided with MJPEG decoder release package, but we are seeing the issue. Can you help us to reproduce it.

We wanted know which environment corruption is visible? in system along with JPEG decoder is anything else running ? like capture or streaming thread which uses shared resources.

Have you tried running in standalone? are you seeing the issue here?

Hi Veeranna,

Thanks for taking the time to investigate the problem.

In my case, I have both JPEG encoder and decoder instantiated sharing the resources (using the same SCRATCH_GROUP_ID). Only one of them runs at a time. The problem exists only when decoding an image of format YUV_420SP.

Let me know if you need more information.

Thanks,

Jimmy

hi, Veeranna.

when will it be released? i saw it's still Version 1.00.00.10 here http://software-dl.ti.com/dsps/dsps_public_sw/codecs/DM36x/index_FDS.html

im looking forward to see it.

regards, Mike

Hello Mike,

Is artifact issue got solved with attached lib in previous post?. after fix we are not seeing the artifacts at our end. If you test with lib and conforms about the fix we will do final release.

Thanks,

Veeranna