ProcessImage w/ decoded source (Custom pixel accessor) about photosauce HOT 7 CLOSED

Good deal. I'm looking to replace my WIC stuff with managed implementations, so I'll have an analogous custom interface, but it'll work the same. I'll keep this on my to-do list.

from photosauce.

That's an interesting one. I've been considering how I can make my processing pipeline more extensible without sacrificing the things that make it fast and lightweight. The pipeline today is set up with a pull model built around IWICBitmapSource. Using that interface allows me to request only the pixels I need immediately (typically enough to process a single output scanline) into a small buffer. That, in turn, allows the upstream pixel source to avoid loading large images in memory all at once. If you've already got the whole image decoded and sitting there, it's easy enough to fill that buffer, but then you end up copying the data when the original could be accessed instead. I'll have to give that some thought.

Oh, and your example method signature leaves out the destination buffer/stream. Was that just an oversight, or were you thinking the input buffer could be recycled to store the output?

from photosauce.

from photosauce.

This is also something that would help us a lot. Currently we can't plug MagicScaler into our pipeline and maintain the performance benefits.

from photosauce.

I've finally got the architecture changes done to make this possible and have taken a stab at the interfaces. Pixels can be fed into or pulled out of the pipeline using the IPixelSource interface, which is a simplified version of IWICBitmapSource. Here's how they go:

Option 1. Feed pixels into MagicScaler pipeline, and complete processing in the pipeline.

var pix = new TestPatternPixelSource(1024, 768, PixelFormats.Bgr24bpp);
using (var fs = new FileStream(@"c:\img\testpattern.png", FileMode.Create))
    MagicImageProcessor.ProcessImage(pix, fs, new ProcessImageSettings { Width = 100, SaveFormat = FileFormat.Png });

Option 2. Use MagicScaler to decode, color correct, resize, and sharpen... then pull the pixels out of the pipeline for continued processing. This example resizes an image and draws it onto a System.Drawing.Bitmap, allowing further manipulation.

using (var pl = MagicImageProcessor.BuildPipeline(@"c:\img\big.jpg", new ProcessImageSettings { Width = 100 }))
using (var fs = new FileStream(@"c:\img\small.png", FileMode.Create))
using (var bmp = new Bitmap(pl.PixelSource.Width, pl.PixelSource.Height, PixelFormat.Format24bppRgb))
{
    var rec = new Rectangle(0, 0, bmp.Width, bmp.Height);
    var lck = bmp.LockBits(rec, ImageLockMode.ReadWrite, bmp.PixelFormat);
    pl.PixelSource.CopyPixels(rec, lck.Stride, (lck.Stride * lck.Height), lck.Scan0);
    bmp.UnlockBits(lck);
    // do some drawing or something
    bmp.Save(fs, ImageFormat.Png);
}

Option 3. You can do IPixelSource both in and out

The IPixelSource provided as input can theoretically use any pixel format supported by WIC, and the output will always be BGRA, BGR, or Grey. Those Guids are included as static fields the PixelFormats class for ease of use.

An alpha build is up on nuget now.

Does that cover your needs or are there some scenarios I've missed?

from photosauce.

This API looks perfect. Going to give it a whirl in the next few hours.

from photosauce.

Works perfectly!

from photosauce.