iMX6Q Jpeg encode

[alexisfronteau]

Hi,

I would like to jpeg encode packed YUV 4:2:2 raw images with an Apalis iMX6Q.

I have tried to use the libimxvpuapi but it seems to work only with planar 4:2:2

images. Are there others libraries that I can use?

I would like to encode at a rate of 160Mpx/s , is it realistic?

Best regards,

Alexis

[art]

Actually, VPU can process only planar image data on its input due to its hardware nature. A raw data should be re-formatted to planar before being processed by VPU.

Yes, it is realistic to encode images at 160Mpix/s.

Have a great day,
Artur

-----------------------------------------------------------------------------------------------------------------------
Note: If this post answers your question, please click the Correct Answer button. Thank you!
-----------------------------------------------------------------------------------------------------------------------

[alexisfronteau]

First, thankes for your reply Artur!

But can I encode images of size 3360x2496? cause in the vpu documentation (http://hands.com/~lkcl/eoma/iMX6/VPU_API_RM_L3.0.35_1.1.0.pdf ) it is written about 1920x1088 images.

[art · Accepted solution]

The Chapter 69 of the i.MX6Q RM says that MJPEG can be encoded at up to 8192x8192 size.

Artur

[alexisfronteau]

Ok,

I find the reference manual but when I try to compress an image of 8Mpx, it takes 200ms using the libimxvpuapi!

How do I need to do to reach the 160Mpx/s?

Alexis

[art]

Do you use the L3.0.35 Linux BSP by FSL/NXP? If so, please note that this BSP is quite early and not well-optimized. The better way is to migrate to most up-to-date BSP release. Current release is L3.14.52.

Also, take into the account other possible overhead, possibly related with image re-formatting, reading/storing etc.

Artur

[alexisfronteau]

I'm currently using the  Linux-3.14.28-V2 linux BSP and i have written a short program just for encode one image.

Alexis

[art]

Seems that you use the outdated VPU API Reference document. Please use the attached one. Also, still take into the account the software overhead, related with image reading/storing, it can be significant.

Best Regards,

Artur

[alexisfronteau]

Thanks for all!

Best Regards,

Alexis

[art]

Sorry, forgot for the attachment. Please find it :-)

Artur