JPEG Decoder

Lucifar He

Expert 1425 points

Hi all,

Recently I found an issue of JPEG decoder.

The decoder works fine when decoding the JPEG I created in Windows,

But some times it does not work when decoding the JPEG encoded by TI's JPEG encoder.

The "process" call returns -1.

Those images seems fine in Windows

Do you have any suggestion to solve the problem?

Best Regards,

Lucifar

over 12 years ago

0 Chad Courtney over 12 years ago

TI__Mastermind 30825 points

Lucifar,

Can you be more specific? What JPEG decoder are you referring to? Are you saying that a TI Source JPEG Decoder is having issues decoding a JPEG image that was created via an TI Source JPEG Encoder?

Can you point to the reference source?

Best Regards,
Chad

0 Lucifar He over 12 years ago in reply to Chad Courtney

Expert 1425 points

Hi Chad,

Yes, I am saying that.

It's C66x jpeg decoder 03.01.00.01 and C66x jpeg encoder 02.02.00.01.

I create a memory for YUV420P, fill Y with random data (from 0-255), and U V with 128, encode it as YUV 420P, and try to decode.

In an endless testing of above process,

the "process" call of the decoder occasionally failed (around 50% of them).

Though the data is random, which means nothing to human, but I think it should not be the reason why the decoder failed,

since when I trying the same procedure with YUV420I format, it works very well.

Here I attached a image that the decoder failed to decode. You can try.

4572.C674_REMI0009.rar

0 Chad Courtney over 12 years ago in reply to Lucifar He

TI__Mastermind 30825 points

Lucifar,

I'm going to ask one of our image processing experts to have a look at this.

Best Regards,

Chad

0 Lucifar He over 12 years ago in reply to Chad Courtney

Expert 1425 points

Thank you chad

0 Paula Carrillo over 12 years ago in reply to Lucifar He

TI__Mastermind 39160 points

Hi Lucifar,

I could reproduce the issue. JPEG Picture header shows interlace as “none” however “start of scan marker” (0xFFDA) is founded 3 times in the picture. For reading the header I used "Image magic" detect tool. As a test, I converted REMI0006.JPG to sequential in order to get rid off of the FF DA. For converting I used "Image magic" convert tool. The result was that the converted JPEG could be decoded without issues. In another test, I converted REMI0006.JPG to JPEG interlaced (using Image magic convert tool) and what I can notice is that now we have 10 scan mark (FF DA) instead of 3 and the JPEG header shows Interlace as "JPEG". So I think it is not a decoder issue but more like a JPEG picture header problem.

Thank you,

Paula