Video capture problems due to driver version?

When you say horizontal lines and motion the first thought that comes to mind is interlacing, however the horizontal lines you have there appear to be several pixels tall, which seems to rule out interlacing. It sounds like the frames are correct in a prior processor and that the horizontal line artifacts only appear after capture by the VPFE, but before any resizing operation, so this would seem to be some sort of digital video format issue or problem with the VPFE. Is the VPFE setup to do any other processing (such as the preview engine) or is it just passing the data from the bus directly into memory? Since you mention another device (I am assuming a processor) that you can check the data from I imagine you have made modifications to the capture driver to handle it?

R&D said:

At certain sizes (when rising above QVGA) we have the stripe effect that can be seen in the next image around the moving hand:

If the issue is visible before the resizer than is the QVGA mentioned here the prior processor sending in a QVGA image instead of a D1 image? I am kind of confused as to when the artifacts exist and do not exist.

R&D said:

Could it be possible that the stripes are caused because of using old driver versions? Does anyone know? Might it be solved by updating the software?

There are newer software versions available however it is hard to say if any of the changes would impact this particular issue without knowing more about what causes it. The new software may be worth a try, though if you have driver customizations already in place it may not be worth the effort.

Yes, the first thing we thought about was interlacing because (I forgot to say it in the first post) we are passing interlaced video (BT.656).

I am not completely sure, but I think we are passing the video directly to the resizer without any other processing in the VPFE. I'll have a look at it.

Aparently the artifacts appear always before the resizer, but they become more visible as we make the resizer output bigger. When this output is QVGA, aparently they are not visible. But as we increase this output size the stripes can be seen. But as I said before, they are also present at the resizer entrance, so it's not due to a resizing artifact.

Besides, to avoid any doubts, we are not changing the size of the original video at the resizer entrance. It is fixed.

R&D said:

Yes, the first thing we thought about was interlacing because (I forgot to say it in the first post) we are passing interlaced video (BT.656).

Unfortunately your horizontal artifacts look too thick for this, an interlacing artifact should show up with a line by line tearing. Is the hand in the image moving very fast or relatively slowly? I ask because I am curious if the artifacts are dependent on a field to field or frame to frame difference in the image or if there is some longer term residual effect going on here which would point to a problem earlier in the signal chain as the VPFE does not keep data around that long.

R&D said:

When this output is QVGA, aparently they are not visible. But as we increase this output size the stripes can be seen. But as I said before, they are also present at the resizer entrance, so it's not due to a resizing artifact.

In other words when you scale down the image the horizontal artifacts seem to blend themselves out of visibility, but when you scale up the image it amplifies the artifacts?

First of all, thanks!

Answering your question about the hand speed, we have seen the stripes when the camera catches a person walking around, so it's not necessary that the object moves too quickly for the stripes to appear.

On the other hand, I think I didn't make myself understood. I'm sorry. We are always downscaling the image in the resizer from a original size of about VGA. When the output image is around QVGA, the artifacts are hardly visible. And they become more visible as we change the image (I mean, the resizer output) to make it bigger, but always below its original size.

bernie,as my project is relaetd to this thread im jus posting my problme here,plz tkae time to suggest me,my project is i have to connect a live camera to the dsp and access data rom pc->dsp via ethernet...i jus started collecting data about how to do this...?? can u please suggest how this process can be done????do i need to write a code in CCS now???can u provide me  some ideas fo this???

Hi Bernie and All,

We are having the same issue with our captured frames.. We are seeing horizontal lines on the raw image in each frame. The frame are captured raw via cmos -> vpfe -> ram-> disk. We are using Aptina sensor mt9v024 and DM6446 with LSP_2_00_00_06.

-clint

Dear clints,

                    plz take time to suggest me,my project is i have to connect a live camera to the dsp and access data rom pc->dsp via ethernet...i jus started collecting data about how to do this...?? can u please suggest how this process can be done????do i need to write a code in CCS now???can u provide me  some ideas fo this???