Re: [AGG] Subpixel problem - 90/180 deg. lines

[Ben H. <be...@im...>]

I think the 'right' thing to do does depend on your entire pipeline. If what you're doing is building a GUI, then maybe pixel snapping at the end does make sense. But then how do you deal with things like curves around the edges of buttons, etc, where you do want sub-pixel positioning?

From: Oyvind Idland [mailto:oyv...@gm...]
Sent: 08 March 2011 01:45 AM
To: Anti-Grain Geometry
Cc: Ben Harper
Subject: Re: [AGG] Subpixel problem - 90/180 deg. lines

Thing is, the ordinates are already piped through other converters, such as parallel shifting (in some cases), conv_transform (always), etc.
I have objects in one coordinate-system affined into pixel-space coordinates.

Making a conv_transform etc. aware of pixel adjustment, sounds quite complicated, not to mention a heck lots of work ?

Correct me if I misunderstand you.


-- Oyvind

On Mon, Mar 7, 2011 at 9:32 PM, Ben Harper <be...@im...<mailto:be...@im...>> wrote:
I think it's the wrong approach to use rounding to fix that. A (0.5,0.5) translation converter is better, since you can then mix and match integer and non-integer coordinates. Even better is if your entire pipeline is aware of the pixel coordinate system, and you apply the layout rounding at the appropriately higher level (perhaps at design time). WPF et al have their 'Layout Rounding' property that you can set on UI objects - I don't know exactly at what stage of the pipeline that gets applied.. all I'm really saying is that it's probably better that you understand the problem fully and apply the correction to the right place of whatever stack it is you're building.

Ben
________________________________
From: Oyvind Idland [oyv...@gm...<mailto:oyv...@gm...>]
Sent: Monday, March 07, 2011 6:12 PM
To: Anti-Grain Geometry
Cc: Phil Endecott
Subject: Re: [AGG] Subpixel problem - 90/180 deg. lines

Yep, thats the problem.

I think I'll make a new conv_round() class and wrap it up that way in those cases.


-- Oyvind
On Mon, Mar 7, 2011 at 3:58 PM, Phil Endecott <spa...@ch...<mailto:spa...@ch...><mailto:spa...@ch...<mailto:spa...@ch...>>> wrote:
Oyvind Idland wrote:
> When a line segment is 90 or 180 degrees (has same x or y), the subpixel
> logic makes the lines appear as 50% transparent and twice as fat as they
> should be.

Presumably that is the standard issue of whether 1-px-wide lines are
better centred on whole coordinates or on 0.5 coordinates, i.e. are
"pixels" points or unit-squares.


Phil.




------------------------------------------------------------------------------
What You Don't Know About Data Connectivity CAN Hurt You
This paper provides an overview of data connectivity, details
its effect on application quality, and explores various alternative
solutions. http://p.sf.net/sfu/progress-d2d
_______________________________________________
Vector-agg-general mailing list
Vec...@li...<mailto:Vec...@li...><mailto:Vec...@li...<mailto:Vec...@li...>>
https://lists.sourceforge.net/lists/listinfo/vector-agg-general



------------------------------------------------------------------------------
What You Don't Know About Data Connectivity CAN Hurt You
This paper provides an overview of data connectivity, details
its effect on application quality, and explores various alternative
solutions. http://p.sf.net/sfu/progress-d2d
_______________________________________________
Vector-agg-general mailing list
Vec...@li...<mailto:Vec...@li...>
https://lists.sourceforge.net/lists/listinfo/vector-agg-general