I am looking to implement a mechanism that combines bitmaps together in a variety of complex ways, using ternary raster operations like you can in Windows.
The idea is to be able to blt an image to a destination using any kind of combination of the source, brush, and destination pixels (source AND destination, source AND brush AND destination, etc.)
This is supported by Windows GDI in what's called Ternary Raster Operations (check out http://msdn.microsoft.com/en-us/library/dd145130(VS.85).aspx). Is it possible that OS X and iOS completely lack this functionality? The only thing I've been able to find are blend modes, but they are not nearly as flexible.
Any ideas?
There are no ternary operators in Quartz or AppKit, and almost certainly not in UIKit, either. All drawing in Quartz-land is from a single source (image, color, gradient, etc.) into a single destination (context).
You can have two source images, one as the “source” and the other as the “pattern”/“brush”. For actual pattern drawing, you can use a CGPattern instead of the second image.
Related
I'd like to stack the images from this map: http://earthquake.usgs.gov/regional/nca/soiltype/map/ from linear projection into a leaflet map. The source tiles are in known but nonstandard zoom levels, and leaflet maps want mercator mercator XYZ tiles. In principle, I know how to do this - I have functions for changing XY coordinates into lat-lng coordinates in the two maps, and I just need to map pixels for the target map in terms of pixels in the source map.
This is unfortunately nontrivial, as the source pixels are spread across hundreds of different image files, and I am trying to put them into hundreds more images. Is there a software package that makes this a little bit more straightforward? If there is no library for dealing with this kind of data, it seems like there really should be...
Postgis has the RT_ST_Transform method, which under the hood uses GdalWarp. So, you have at least these two options. If you use Postgis, you will need to actually register/import the images into Postgis, using raster2pgsql and then call RT_ST_Transform on each one and then dump them out again -- which could be scripted to some extent using plpgsql (Postgres's scripting language). There is something of a learning curve involved with using Postgis raster, which may be worthwhile if you plan to do any other image processing analysis. You could also write a shell script (or similar) to automate gdalwarp if you don't wish to go the Postgis route.
For a less formal method than gdalwarp (an excellent program), you can check out the Leaflet plugin Leaflet.imageTransform, that can transform and image on the fly in the browser.
What's the easiest way to create a custom blendmode for Adobe Photoshop?
I want to be able to blend two images together according to rules that can't be created by combining existing blendmodes
(i.e. the blendresult shall be dst = backLayer + (frontlayer*2-1) which can't be achieved by applying linear dodge twice and then subtracting a white layer since clamping will occur - and when working in a 32bit workspace, the blendmodes will not behave as expected anymore)
I tried to program pbk kernels in PixelBender, but Photoshop's PixelBender doesn't seem to support pbk kernels that take more than one image as input (can work only as filter on a single image).
What's the most straight forward way to do this?
Check if the following technique works:
http://www.photoshopgurus.com/forum/general-photoshop-board/25925-custom-blending-modes-possible.html
I have the layout of a website in CorelDraw X4 and I need to move it to Fireworks CS5 (for many reasons). The thing is that, apparently, the only method I was able to find on the Internet doesn't work very well. What I do is to export the file from Draw to AI (Adobe Illustrator) format. Then I import the file in fireworks, but there, strange things happen. The first thing is that borders are thicker after this process (1 to 4) but the real problem comes with some objects thar are converted to bitmaps (or so I think). When I delete all the bitmaps, only a few objects remain and that's obviously undesired. In my original file I use transparencies and gradients applied to many different objects.
Do you know why this happens and/or a possible solution? Thanks!
Edit: I think I'm getting closer! Apparently AI format doesn't support transparencies, so... I get all trasparencies out before exporting (not very nice, but what can I do, right?) or I ungroup all objects once imported into Fireworks and then carefully delete the bitmaps (which seem to be the approximation of transparencies for AI). All this is just about testing, if someone knows what happens or of another solutions, please, thow light. Thanks...
Ok, as nobody else answered my question I suppose I can consider myself capable to provide more information than anybody else, ha!
I've been studing the case and reached to a semi-solution. Apparently, AI is the only format supporting vectors that can be exported and imported by both editors. The problem with this is that AI doesn't support transparencies nor shadows. So... If you really want to do this, be prepared to work a bit.
What I did was to copy all the shapes without effects using this export/import method (surprisingly, line thickness was preserved correctly this time), then I examined shape by shape in Corel and applied the same (or its best aproximation) effects in Fireworks. This wasn't easy because the way both programs apply shadows and transparencies is a bit different. Yeah, it's not easy, but it's all we got...
Little tip: In my case I had some shapes with transparencies AND shadows. In Corel these shadows where strong as if the object was solid (not transparent). In fireworks, the shadow disappears with the object when the transparency is applied (as logically expected). What I did to solve this was to copy the object and apply a Gaussian blur to the object in the back, acting as a full shadow even when the object in the front was fading to transparent.
In my iPhone App I'd like to draw a few formulas. How can I manage that with quartz 2d? Is there a way to build formulas like for example in latex? Or are there any existing frameworks? Thanks.
As the developer of an iPhone application which does just that, trust me when I say typesetting equations is not a trivial undertaking. In my case, I used Core Animation layers to construct the sub-elements of a parsed equation. The equations are constructed hierarchically, and the operations that compose them are laid out as such. Each operation is contained within its parent operation's layer, and laid out following the rules of that particular operation.
For rendering the visual elements of the operations within an equation, I used Quartz to draw lines, symbols, etc., but most of the drawing was simply text drawn within a CALayer using the NSString text drawing extensions.
I did override the standard CALayer rendering architecture for the generation of PDFs from these equations, because CALayers don't render as vector elements by default. For an example of how this rendering works in my application, see the open source Core Plot project, which does the same thing at its base level.
I do output to LaTeX from the equations, which is pretty simple once you've parsed them into a hierarchical data structure, but parsing them from LaTeX into that structure is proving a little trickier.
For simple text equation input and evaluation, you might find Graham Cox's GCMathParser to be of use.
There's no real Tex engine implemented for Cocoa touch that I have heard of, but there is a lightweight Javascript formulae-layout engine: jsMath.
How would I smooth a line (UIBeizerPath) or a set of points? Right now it draws it jagged. I read about spline interpolation, could anyone point me to an implementation of this in cocoa or C or give me an alternate line smoothing algorithm.
I don't think you need to do Bezier paths with curves. You can keep drawing straight line segments but add more data points with interpolation. This is especially important because I'm assuming you want to smooth only on one axis so you don't end up with odd things like loops in your graph.
So you want to add more points to your source data, between the existing points, and use an interpolation algorithm that's more sophisticated than a linear interpolation. There are many to choose from. Quadratic? Sine-based? Many, and it depends on what kind of data you're using.
Quartz (which UIKit uses for drawing, and in many places makes you use directly for drawing) has anti-aliasing support built-in. Most contexts have it turned on already, so you should not have aliased (jagged) drawing unless you're turning anti-aliasing off. So, stop doing that. :-)
The contexts that don't have it turned on by default are mostly those where it isn't appropriate, such as PDF contexts and CGLayer contexts. The documentation implies that those contexts don't even support anti-aliasing, which makes some amount of sense.
CGContext provides a couple of functions for turning anti-aliasing on and off, but you should never need to call them except when you want aliasing, which you don't. You could try turning it on using those functions; if that works, then you should investigate why it was ever off in the first place.
Are you drawing the path from within a CALayer? That may be why it's off; there's an Info.plist key you have to turn on to get anti-aliasing turned on by default in such contexts.
I've found that if you draw a line or an image on the edge of your frame that it will appear jagged. Move the line in a few (or grow your frame) and it should appear nice and crisp. Again, not sure if that's your question or not but it has bit me a few times.
For instance if you are displaying an image inside a CALayer, make sure there is space between the image and the frame if you are doing anything but 90 degree angles.