mikeash.com pyblog/friday-qa-2012-08-31-obtaining-and-interpreting-image-data.html comments

starmessage - 2014-08-12 19:49:58

Tue, 12 Aug 2014 19:49:58 GMT

Very good description.

I am trying to write a cross platform (windows + OSX) image class and have difficulties in the MAC part.
I am using NSImage as the basic element and the bmp representation to draw pixels in it.
While this works ok when I load an image from a file into NSimage and then there is a ready bmp representation to modify its pixels before I draw the nsimage on an nsview.

My problems begin when I want to create a blank nsimage and modify its pixels.
I am failing at all different sequences of attaching a bmp rep ont the image.

I will examine your proposals.

mikeash - 2012-09-05 14:21:33

Wed, 05 Sep 2012 14:21:33 GMT

Johan Sørensen: Thanks, that was silly of me. Fixed now.

Johan Sørensen - 2012-09-05 08:56:06

Wed, 05 Sep 2012 08:56:06 GMT

Nice introduction!

Your NSBitmapImageRep initializer is missing the bitsPerSample argument though

Josh Bleecher Snyder - 2012-09-01 14:40:53

Sat, 01 Sep 2012 14:40:53 GMT

bytesPerRow and friends are used not just for memory alignment but also for copy-free subimages. It's a significant feature and design consideration that's worth going into.

mikeash - 2012-09-01 00:05:37

Sat, 01 Sep 2012 00:05:37 GMT

Or if you're rendering manually somewhere, you can just set a transform before you render.

rudy - 2012-08-31 22:44:06

Fri, 31 Aug 2012 22:44:06 GMT

@Zev

presumably you're using the image you've gotten from your CGBitmapContext in a UIView or on a CALayer? you can just apply an affine transform to the view/layer and it'll be rotated for you.

Zev Eisenberg - 2012-08-31 21:45:40

Fri, 31 Aug 2012 21:45:40 GMT

Marginally related question: on iOS, if I have some pixels rendered into a CGBitmapContext, if I want to render a rotated version of that image, do I have to grab an image and draw it into a new context with a transform applied? Or is there a quicker way?

mikeash - 2012-08-31 16:24:03

Fri, 31 Aug 2012 16:24:03 GMT

The nice thing is that once you understand the basics of this stuff, using CGBitmapContext is pretty easy.

Torsten - 2012-08-31 16:07:59

Fri, 31 Aug 2012 16:07:59 GMT

A nice introduction! A frequent use for this technology is to load textures into OpenGL. The advantage is that any format can be used as a texture, including PSDs and even PDF files. That means no conversion step when testing out new art.

(Personally, though, I prefer to use CGImage and CGContext for this. It takes a bit more work, but one has more control especially over where the Alpha value ends up)