DRX Artifacts

[Irri]

It seems to work very well, but when I used it on full with this image I got a strange white line around an edge:

 

[Dan Hudgins]

White line around edge...

White line around edge...

It seems to work very well, but when I used it on full with this image I got a strange white line around an edge:

Are you talking about the white line around the gray rectangle in the image on the right?

It looks like a sharpen artifact, turn off any sharpen value. Sometimes compressors sharpen images, so that can show up in compressed images.

Do not use "unsharp mask" it is like sharpen.

 

[Irri]

Yeah, the white line.

They are both direct from RedAlert. The only difference is that the right has DRX.

 

[Dan Hudgins]

Luma artifact?

Luma artifact?

Yeah, the white line.

They are both direct from RedAlert. The only difference is that the right has DRX.

See this link for info about DRX,

http://www.reduser.net/forum/showthread.php?t=4187&page=4

From what I can make out, DRX adjustes the luma vs. the chroma when one or two of the Bayer colors is clipped to keep tones near white white, so because the de-Bayer works like a sharpen or unsharp mask by mixing +/- values of the pixels around the pixel in the center with off color, you get an unbalance in the de-Bayer that may look like a sharpen artifact?

Without the DRX details that seems to be what is going on, an unbalance in the de-Bayer due to clipping, try reducing the exposure to avoid clipping and turn the DRX off?

 

[Irri]

Thanks Dan,

It's not an image I am trying to save. It's a completely blown out greyscale. I was just interested in how DRX works.

 

[Graeme Nattress]

DRX doesn't look at any adjacent pixels - it only corrects colours.

Graeme

 

[Dan Hudgins]

What's making the white outline?

What's making the white outline?

DRX doesn't look at any adjacent pixels - it only corrects colours.

Graeme

When you do the de-Bayer the pixels around the pixel and have some that are clipped, so you need to adjust the de-Bayer to correct the color shift with negative values on the higher pixels, in essence an negative mask like an unsharp mask since you cannot correct the color of a single pixel since single pixels color is not defined until after the de-Bayer. If you are just clipping the high pixels before de-Bayer, then some colors would go off balance that are made using that primary?

Single pixels do not have a color untill the pixels around them are interpolated and the chroma is adjusted by mixing the adjacent colors, of which one or two are off because of clipping.

DRX does not seem to extend the color brightness range, rather it seems to clip the unclipped colors to match the lowest one?

So if you convert the DRX image to monochrome, you would have less range than without DRX on, since one of the off colors would still record brightness?

Anyway, if when the DRX is on he gets a white outline border, and not when it is off, then what is causing the white border if not the DRX?

 

[Graeme Nattress]

I can see the line in the original image before DRX applied though.

DRX does indeed extend the brightness range, although that is very lighting dependent on what benefits it brings. It doesn't clip to match the lowest, but fill in the lowest clip to match the least clipped.

Graeme

 

[Damon Meledones]

Perhaps the line is a result of a subtle unsharp mask having been applied to the screen used to print the chart?

 

[Irri]

The patches on the chart are slightly raised so they may have got a slight ping from the lights.