--- title: "Extract images from a PDF in C#" url: https://recursiveai.org/onepdf/cookbook/extract-images/ description: "Extract images from a PDF in C#: JPEGs byte for byte, everything else as PNG with transparency, JPEG 2000 and CMYK included. Runnable code." version: "1.15.0" --- # Extract images from a PDF in C# Run each page through a small class that overrides `DrawImage`: OnePDF calls it once for every image the page draws. A JPEG in a PDF is a complete JPEG file, so its bytes can be saved as they are; every other image is raw pixels, which the program writes as PNG with OneImage. It runs on a USGS fact sheet, a public PDF with JPEG 2000 photos, CMYK maps and transparent icons. ```shell dotnet new console -o PdfImages && cd PdfImages dotnet add package RecursiveAI.OnePdf --version 1.15.0 dotnet add package RecursiveAI.OneImage.Pdf --version 1.10.4 curl -LO https://pubs.usgs.gov/fs/2016/3020/fs20163020.pdf ``` Put this in `Program.cs` and run `dotnet run`: ```csharp using System.Runtime.InteropServices; using OneImage; using OneImage.Pdf; using OnePdf; // OnePDF decodes Flate, LZW, CCITT and the other PDF filters. // This adds JPEG, JPEG 2000 and JBIG2. PdfImageCodecs.Register(PdfImageDecoderRegistry.Default); using PdfDocument doc = PdfDocument.Open("fs20163020.pdf"); var saver = new ImageSaver("images"); foreach (PdfPage page in doc.Pages) { saver.PageNumber = page.Index + 1; // runs the page and calls DrawImage for every image it draws page.Interpret(saver); } Console.WriteLine($"{saver.Jpegs} saved as .jpg (original bytes), " + $"{saver.Pngs} as .png, {saver.Skipped} skipped"); sealed class ImageSaver(string folder) : PdfDevice { // an image drawn on many pages is saved once readonly HashSet _seen = []; public int PageNumber { get; set; } public int Jpegs { get; private set; } public int Pngs { get; private set; } public int Skipped { get; private set; } public override void DrawImage(in PdfImage image, in PdfGraphicsState state) { if (image.Stream is { } stream && !_seen.Add(stream)) return; var encoded = new PdfImageOptions { PreferEncoded = true }; PdfImageData? data = image.GetData(encoded); // a stencil is a shape painted in one color if (data is null || data.IsStencil) return; Directory.CreateDirectory(folder); int n = Jpegs + Pngs + Skipped + 1; string path = Path.Combine(folder, $"page{PageNumber}-{n}"); // A gray or RGB JPEG with no mask and no /Decode is a complete JPEG file. if (data.EncodedFilter == "DCTDecode" && data.Alpha is null && data.Semantics.Decode is null && data.Semantics.ColorSpace?.ComponentCount is 1 or 3) { File.WriteAllBytes(path + ".jpg", data.EncodedData.ToArray()); Jpegs++; return; } // decoded: 8-bit samples in their final color space, mask as alpha data = image.GetData(); if (data is not { IsDecoded: true } || data.ColorSpace!.Kind == PdfColorSpaceKind.Lab) { Skipped++; return; } ToImage(data).Save(path + ".png"); Pngs++; } static Image ToImage(PdfImageData data) { ReadOnlySpan s = data.Samples.Span; int w = data.Width, h = data.Height; PdfColorSpace cs = data.ColorSpace!; Image img = cs.ComponentCount switch { 1 => Image.FromPixels(MemoryMarshal.Cast(s), w, h), 3 => Image.FromPixels(MemoryMarshal.Cast(s), w, h), _ => Image.FromPixels(MemoryMarshal.Cast(s), w, h), }; if (!cs.IccProfile.IsEmpty) // keep its profile img = img.AssignProfile(ColorProfile.FromIcc(cs.IccProfile.Span)); if (cs.ComponentCount == 4) img = img.ConvertColor(ColorProfile.Srgb); // PNG has no CMYK if (data.Alpha is { } alpha) { GrayAlpha8[] values = [.. alpha.Values.ToArray().Select(a => new GrayAlpha8(0, a))]; var mask = Image.FromPixels(values, alpha.Width, alpha.Height); if (alpha.Width != w || alpha.Height != h) // a mask may have its own resolution mask = mask.Resize(w, h, Fit.Stretch); img = img.AddAlpha().Composite(mask, alpha: AlphaComposition.DestIn); } return img; } } ``` ```text 5 saved as .jpg (original bytes), 38 as .png, 0 skipped ``` The `images` folder now holds every distinct image of the six pages. Poppler’s `pdfimages -list` counts the same 43 image objects in the file, plus the soft masks that the program applies as transparency instead of saving separately. ## What each part handles - **Duplicates.** `image.Stream` is the image object, so a logo drawn on every page is saved once. Inline images have no stream and are saved each time. - **JPEGs as they are.** The bytes are only a faithful file when the PDF adds nothing on top: no mask, no `/Decode` array, gray or RGB. CMYK JPEGs can carry an inverting `/Decode`, so they go through the decoder. - **Decoded pixels.** `GetData()` applies every PDF rule first: palettes, spot colors, `/Decode` arrays and color-key masks. - **Color.** An ICC-based image keeps its profile in the PNG. CMYK is converted to sRGB. Lab images are skipped here. - **Transparency.** `data.Alpha` is the soft mask, color-key mask or JPEG 2000 alpha, stretched to the image’s size when it has its own resolution. ## Limits - **Images, not pages.** This saves the pictures embedded in the file. To get a picture of a whole page, with its text and drawings, render it with [OneImage](https://recursiveai.org/oneimage/). - **Codecs.** Without `RecursiveAI.OneImage.Pdf`, OnePDF still returns JPEG bytes and decodes Flate and the other PDF filters, but JPEG 2000 and JBIG2 images come back undecoded, and you need your own PNG encoder. - **Size limits.** An image’s declared size is checked against `PdfLimits` before any memory is allocated, and `GetData` throws `PdfLimitException` above it. That matters for PDFs users upload. - **Stencil masks** (one-bit shapes painted in a single color, such as the black layer of some scans) are skipped. To put an image into a PDF instead, see [Add an image to a PDF](https://recursiveai.org/onepdf/cookbook/add-image-to-pdf/). The drawing calls a device receives are in [Display-list text format](https://recursiveai.org/onepdf/docs/display-list-format/).