High performing Bitmap Drawing solution in c# - c#

I have a use case where I need to render a collage of bitmaps as a preview. The application is implemented as a MVC based REST service, and I have a fairly vanilla implementation:
using (var bitmap = new Bitmap((int)maxWidth, (int)maxHeight))
{
using (var graphic = Graphics.FromImage(bitmap))
{
graphic.InterpolationMode = InterpolationMode.HighQualityBicubic;
// now again for each mod
foreach (var mod in mods)
{
var frame = frames.First(f => f.Index == mod.FrameIndex);
var fileInfo = GetFileInfo(mod);
using (var modImage = Image.FromFile(fileInfo.FullName))
{
graphic.DrawImage(modImage, (int)frame.Left, (int)frame.Top);
}
}
bitmap.Save(previewFileName);
}
}
While this code works fine, it performs very poorly (especially with larger images). I am open to using third party libraries as well, I just need a faster performing solution.
Any help would be mostly appreciated.
Update
To clarify the use case, caching doesn't help. These images are uploaded by the customer, then they request a preview of the selected collage. It's the writing of the images to the collage that is slow.

I just now realized that you're drawing the modImages in their original size (because you're using the override that only takes a left and top coordinate). If that's what you want, you could as well just use the DrawImageUnscaled()-Method of the graphics class, that should be much faster:
var frame = frames.First(f => f.Index == mod.FrameIndex);
var fileInfo = GetFileInfo(mod);
using (var modImage = Image.FromFile(fileInfo.FullName))
{
graphic.DrawImageUnscaled(modImage, (int)frame.Left, (int)frame.Top);
}

How important is the image quality? If you need faster results, you should try InterpolationMode.NearestNeighbour, this should be much faster, but results will be rather low-quality.
HeighQualityBicubic (what you're currently using) produces best results, but at lowest performance.

If you have a static number of Thumbnails, similar to Netflix, I would create a Cache in memory that keeps the thumbnail bitmaps. This could even be a static object. Once you have generated the thumbnail, you never need to generate it again. It would be a cleaner approach than trying to find a faster hammer to hit it with every time.
Example :
if(MyThumbnailImageDictionary["whateverimageId"] != null)
return (Bitmap)MyThumbnailImageDictionary["whateverimageId"];
If the thumbnails are dynamic such as from a user uploading random pictures, then this method will not help as the size of the image cache will grow tremendous.
You can use the following code to generate a Thubnail, but I am not sure how much faster it would be :
Image image = Image.FromFile(fileName);
Image thumb = image.GetThumbnailImage(120, 120, ()=>false, IntPtr.Zero);

Related

The best practice to avoid out of memory exception

I need to convert tif to webp and I am using wep wrapper class from this project - https://github.com/JosePineiro/WebP-wrapper/blob/master/WebPTest/WebPWrapper.cs
If I use regular for loop everything seems to be working correctly but I need to sped up the process and that is why I am using parallel for each to execute this method. I am getting out of memory exception in some cases. How can I avoid that? I was reading that I can use long memorySize = currentProcess.PrivateMemorySize64
Is this the best practice? I have never used this approach.
And I am looping it like:
Parallel.ForEach(MyList, (row) =>
{
byte[] rawWebP;
Image image = Image.FromFile(row.linkToImage);
Bitmap bmp = (Bitmap)image;
using (WebP webp = new WebP())
rawWebP = webp.EncodeLossless(bmp, row.linkToImage);
});

Resize image on the fly in .net and c#

I'm looking for a way to resize images without saving them on the server. The ways that i have found includes a controller file and such.
Is there a way to get the image from the stream, resize it and add it to the response?
Check out ImageResizer - it's a suite of NuGet packages designed for this exact purpose.
It runs eBay in Denmark, MSN Olympics, and a few other big sites.
Dynamic image processing can be done safely and efficiently, but not in a sane amount of code. It's trickier than it appears.
I wouldn't recommend this but you can do next thing:
using (Image img = Image.FromStream(originalImage))
{
using (Bitmap bitmap = new Bitmap(img, width, height))
{
bitmap.Save(outputStream, ImageFormat.Jpeg);
}
}
Be aware that this could cause OutOfMemoryException.

OutOfMemory Exception when loading lots of Images from Isolated storage

EDIT: I keep getting OutOfMemoryException was unhandled,
I think it's how I am saving the image to isolated storage ,I think this is where I can solve my problem how do I reduce the size of the image before I save it? (added code where I save Image)
I am opening images from Isolated storage sometimes over 100 images and I want to loop over them images but I get a OutOfMemory Exception when there is around 100 to 150 images loaded in to a storyboard. How can I handle this exception, I have already brought down the resolution of the images. How can I handle this exception and stop my app from crashing?
I get the exception at this line here
image.SetSource(isStoreTwo.OpenFile(projectFolder + "\\MyImage" + i + ".jpg", FileMode.Open, FileAccess.Read));//images from isolated storage
here's my code
private void OnLoaded(object sender, RoutedEventArgs e)
{
IsolatedStorageFile isStoreTwo = IsolatedStorageFile.GetUserStoreForApplication();
try
{
storyboard = new Storyboard
{
//RepeatBehavior = RepeatBehavior.Forever
};
var animation = new ObjectAnimationUsingKeyFrames();
Storyboard.SetTarget(animation, projectImage);
Storyboard.SetTargetProperty(animation, new PropertyPath("Source"));
storyboard.Children.Add(animation);
for (int i = 1; i <= savedCounter; i++)
{
BitmapImage image = new BitmapImage();
image.SetSource(isStoreTwo.OpenFile(projectFolder + "\\MyImage" + i + ".jpg", FileMode.Open, FileAccess.Read));//images from isolated storage
var keyframe = new DiscreteObjectKeyFrame
{
KeyTime = KeyTime.FromTimeSpan(TimeSpan.FromMilliseconds(100 * i)),
Value = image
};
animation.KeyFrames.Add(keyframe);
}
}
catch (OutOfMemoryException exc)
{
//throw;
}
Resources.Add("ProjectStoryBoard", storyboard);
storyboard.Begin();
}
EDIT This is how I am saving the image to Isolated storage, I think this is where I can solve my problem, How do I reduce the size of the image when saving it to isolated storage?
void cam_CaptureImageAvailable(object sender, Microsoft.Devices.ContentReadyEventArgs e)
{
string fileName = folderName+"\\MyImage" + savedCounter + ".jpg";
try
{
// Save picture to the library camera roll.
//library.SavePictureToCameraRoll(fileName, e.ImageStream);
// Set the position of the stream back to start
e.ImageStream.Seek(0, SeekOrigin.Begin);
// Save picture as JPEG to isolated storage.
using (IsolatedStorageFile isStore = IsolatedStorageFile.GetUserStoreForApplication())
{
using (IsolatedStorageFileStream targetStream = isStore.OpenFile(fileName, FileMode.Create, FileAccess.Write))
{
// Initialize the buffer for 4KB disk pages.
byte[] readBuffer = new byte[4096];
int bytesRead = -1;
// Copy the image to isolated storage.
while ((bytesRead = e.ImageStream.Read(readBuffer, 0, readBuffer.Length)) > 0)
{
targetStream.Write(readBuffer, 0, bytesRead);
}
}
}
}
finally
{
// Close image stream
e.ImageStream.Close();
}
}
I would appreciate if you could help me thanks.
It doesn't matter how large your images are on disk because when you load them into memory they're going to be uncompressed. The memory required for the image will be approximately (stride * height). stride is width * bitsPerPixel)/8, and then rounded up to the next multiple of 4 bytes. So an image that's 1024x768 and 24 bits per pixel will take up about 2.25 MB.
You should figure out how large your images are, uncompressed, and use that number to determine the memory requirements.
You are getting the OutOfMemory Exception because you are storing all the images in memory at the same time in order to create your StoryBoard. I don't think you will be able to overcome the uncompressed bitmap size that the images require to be displayed on screen.
So to get past this we must think about your goal rather than trying to fix the error. If your goal is to show a new image in sequence every X ms then you have a few options.
Keep using StoryBoards but chain them using the OnCompleted event. This way you don't have to create them all at once but can just generate the next few. It might not be fast enough though if you're changing images every 100ms.
Use CompositionTarget.Rendering as mentioned in my answer here. This would probably take the least amount of memory if you just preload the next one (as opposed to having them all preloaded as your current solution does). You'd need to manually check the elapsed time though.
Rethink what you're doing. If you state what you are going after people might have more alternatives.
To answer the edit at the top of your post, try ImageResizer. There's a NuGet package, and a HanselBlog episode on it. Obviously , this is Asp.Net based, but I'm sure you could butcher it to work in your scenario.
Tackling these kind of problems at design layer usually works better.
Making application smart about the running environment via some configurations makes your application more robust. For example you can define some variables like image size, image count, image quality... based on available memory and set these variables at run-time in your App. So your application always works; fast on high memory machines and slow on low memory ones; but never crash. (Don't believe working in managed environment means no worry about the environment... Design always matters)
Also there are some known design patterns like Lazy Loading you can benefit from.
I don't know about windows phone in particular, but in .net winforms, you need to use a separate thread when doing a long-running task. Are you using a BackgroundWorker or equivalent? The finalizer thread can become blocked, which will prevent the resources for the images from being disposed. Using a separate thread from the UI thread will allow will allow the Dispose method to be run automatically.
Ok, an image (1024x768) has at least a memsize of 3 mb (argb)
Don't know how ObjectAnimationUsingKeyFrames works internal. Maybe you can force the gc by destroying the instances of BitmapImage (and KeyFrames) without loss of its data in the animation.
(not possible, see comments!)
Based on one of your comments, you are building a Time Lapse app. Commercial time-lapse apps for WP7 compress the images to video, not stills. e.g. Time Lapse Pro
The whole point of video playback is to reduce similar, or time-related, images to highly compressed stream that do not require massive amounts of memory to play back.
If you can add the ability to encode to video, in your app, you will avoid the problem of trying to emulate a video player (using 100s of single full-resolution frames as a flick-book).
Processing the images into video server-side may be another option (but not as friendly as in-camera).

WinRT image handling

A friend and I spent the better part of last night nearly tearing our hair out trying to work with some images in a metro app. We got images into the app with the share charm, and then I wanted to do some other work with them, cropping the images and saving them back into the appdata folder. This proved extremely frustrating.
My question, at the end of all this, is going to be "What's the proper way of doing this, without feeling like I'm hammering together a bunch of mismatched jigsaw puzzle pieces?"
When sharing multiple images with the app, they come in as a list of Windows.Storage.StorageFiles. Here's some code used to handle that.
var storageItems = await _shareOperation.Data.GetStorageItemsAsync();
foreach (StorageFile item in storageItems)
{
var stream = await item.OpenReadAsync();
var properties = await item.Properties.GetImagePropertiesAsync();
var image = new WriteableBitmap((Int32)properties.Width, (Int32)properties.Height);
image.SetSource(stream);
images.Add(image);
}
Some searching online has indicated that currently, a Windows.UI.Xaml.Media.Imaging.WriteableBitmap is the only thing capable of letting you access the pixel data in the image. This question includes a helpful answer full of extension methods for saving images to a file, so we used those.
Our problems were the worst when I tried opening the files again later. I did something similar to before:
var files = await ApplicationData.Current.LocalFolder.GetFilesAsync();
foreach (var file in files)
{
var fileStream = await file.OpenReadAsync();
var properties = await file.Properties.GetImagePropertiesAsync();
var bitmap = new WriteableBitmap((Int32)properties.Width, (Int32)properties.Height);
bitmap.SetSource(fileStream);
System.IO.Stream stream = bitmap.PixelBuffer.AsStream();
Here comes a problem. How long is this stream, if I want the bytes out of it?
// CRASH! Length isn't supported on an IRandomAccessStream.
var pixels = new byte[fileStream.Length];
Ok try again.
var pixels = new byte[stream.Length];
This works, except... if the image is compressed, the stream is shorter than you would expect, so you will eventually get an out of bounds exception. For now pretend it's an uncompressed bitmap.
await _stream.ReadAsync(pixels, 0, pixels.Length);
Well guess what. Even though I said bitmap.SetSource(fileStream); in order to read in the data, my byte array is still full of zeroes. I have no idea why. If I pass this same bitmap into a my UI through the sample data group, the image shows up just fine. So it has clearly got the pixel data in that bitmap somewhere, but I can't read it out of bitmap.PixelBuffer? Why not?
Finally, here's what ended up actually working.
var decoder = await BitmapDecoder.CreateAsync(BitmapDecoder.PngDecoderId, fileStream);
var data = await decoder.GetPixelDataAsync();
var bytes = data.DetachPixelData();
/* process my data, finally */
} // end of that foreach I started a while ago
So now I have by image data, but I still have a big problem. In order to do anything with it, I have to make assumptions about its format. I have no idea whether it's rgba, rgb, abgr, bgra, whatever they can be. If I guess wrong my processing just fails. I've had dozens of test runs spit out zero byte and corrupted images, upside down images (???), wrong colors, etc. I would have expected to find some of this info in the properties that I got from calling await file.Properties.GetImagePropertiesAsync();, but no luck. That only contains the image width and height, plus some other useless things. Minimal documentation here.
So, why is this process so painful? Is this just reflecting the immaturity of the libraries right now, and can I expect it to get better? Or is there already some standard way of doing this? I wish it were as easy as in System.Drawing. That gave you all the data you ever needed, and happily loaded any image type correctly, without making you deal with streams yourself.
From what I have seen - when you are planning on loading the WriteableBitmap with a stream - you don't need to check the image dimensions - just do new WriteableBitmap(1,1), then call SetSource().
Not sure why you were thinking var pixels = new byte[fileStream.Length]; would work, since the fileStream has the compressed image bytes and not a pixel array.
You might need to seek to the beginning of the stream to get the pixels array:
var pixelStream = pixelBuffer.AsStream();
var bytes = new byte[this.pixelStream.Length];
this.pixelStream.Seek(0, SeekOrigin.Begin);
this.pixelStream.Read(bytes, 0, Bytes.Length);
I had started working on a WinRT port of WriteableBitmapEx - maybe it could help you: http://bit.ly/WriteableBitmapExWinRT. I have not tested it well and it is based on an older version of WBX, but it is fairly complete in terms of feature support. Might be a tad slower than it is possible too.

C# mvc image upload resizing server side

I've got an webapplication where users can upload images. The current problem i'm running into is that the images being uploaded are being saved to the database in the original format. This causes a lot of performance issues when the images are used on a webpage. I used dotTrace to profile the application and I see significant problems when images are processed from the database.
The idea I have is to resize the image when it's uploaded to the server. Take the following example which I want the application to do when the user uploads a new image;
User uploads an image
The image is being resized to an size of 7.500 x 7.500 in pixels in 72 dpi
The image is being saved into the database
Original file gets disposed
The only stored image is the one mentioned above and the webapplication contains technology to resize this on the fly.
I've already read several topics here on SO. And most of them point me into the direction of ImageMagick. This tool is already familiar at my company, and being used in PHP projects. But are there any good and stable released C# wrappers for this tool? I already found the tools below but they're either in Béta release, Alpha Release or currently not updated.
ImageMagick.NET
ImageMagick APP
I also found this topic on SO. In this topic the following code example is supplied;
private static Image CreateReducedImage(Image imgOrig, Size newSize)
{
var newBm = new Bitmap(newSize.Width, newSize.Height);
using (var newGrapics = Graphics.FromImage(newBm))
{
newGrapics.CompositingQuality = CompositingQuality.HighSpeed;
newGrapics.SmoothingMode = SmoothingMode.HighSpeed;
newGrapics.InterpolationMode = InterpolationMode.HighQualityBicubic;
newGrapics.DrawImage(imgOrig, new Rectangle(0, 0, newSize.Width, newSize.Height));
}
return newBm;
}
In short the questions i have;
Are there any advantages in relation to performance using the example above?
Is there a good and reliable C# wrapper for ImageMagick i can use to do this?
Any other good tips relating to the performance are welcome!
We use the latter approach - I can't comment on performance but it certainly makes handling dependencies simpler.
However, one thing to note is that the above code is probably too simple if your users are able to upload images in all sorts of formats. The underlying library (GDI+) has issues with a lot of color formats, but it also is dependent on the OS version. Here's the core of the code we use:
// GDI+ has problems with lots of image formats, and it also chokes on unknown ones (like CMYK).
// Therefore, we're going to take a whitelist approach.
// see http://bmpinroad.blogspot.com/2006/04/file-formats-pixel-formats.html
// also see http://social.msdn.microsoft.com/Forums/en-US/winforms/thread/c626a478-e5ef-4a5e-9a73-599b3b7a6ecc
PixelFormat format = originalImage.PixelFormat;
if (format == PixelFormat.Format16bppArgb1555 ||
format == PixelFormat.Format64bppArgb)
{
// try to preserve transparency
format = PixelFormat.Format32bppArgb;
}
else if (format == PixelFormat.Format64bppPArgb)
{
// try to preserve pre-multiplied transparency
format = PixelFormat.Format32bppPArgb;
}
else if (format != PixelFormat.Format24bppRgb && format != PixelFormat.Format32bppRgb)
{
format = PixelFormat.Format24bppRgb;
}
// GIF saving is probably still an issue. If we ever need to tackle it, see the following:
// http://support.microsoft.com/kb/319061
// http://www.bobpowell.net/giftransparency.htm
// http://support.microsoft.com/kb/318343
using (Bitmap newImage = new Bitmap(newSize.Width, newSize.Height, format))
{
using (Graphics Canvas = Graphics.FromImage(newImage))
{
using (ImageAttributes attr = new ImageAttributes())
{
attr.SetWrapMode(WrapMode.TileFlipXY);
Canvas.SmoothingMode = SmoothingMode.AntiAlias;
Canvas.InterpolationMode = InterpolationMode.HighQualityBicubic;
Canvas.PixelOffsetMode = PixelOffsetMode.HighQuality;
Canvas.DrawImage(originalImage, new Rectangle(new Point(0, 0), newSize), srcRect.X, srcRect.Y, srcRect.Width, srcRect.Height, GraphicsUnit.Pixel, attr);
newImage.Save(outputImageStream, originalImage.RawFormat);
}
}
}
I've never used ImageMagic, but I have used the GDI+ image resize functions, including on a site that generates and resizes 100,000+ images a day without performance issues.
I would say using the GDI+ methods are just fine. Don't worry about wrapping an external tool or framework.
I've used ImageGen in Umbraco sites. (It's certainly not tied to Umbraco, it's good for any ASP.NET app, it just so happened that some of the Umbraco packages I was using required it.) It's simple to use, and you might be able to get away with the free version...

Categories