Create Image Appearance for Digital Signature

A signature appearance can carry two images alongside its text: Image, typically a scanned handwritten signature placed next to the lines, and WatermarkImage, drawn across the whole box behind everything else. Both are plain Go image.Image values, so anything you can decode you can use.

The split is always into equal halves; there is no way to give the image a third of the box. With no text lines the position is ignored and the image gets the whole rectangle.

Adding the images

imgFile, err := os.Open(imageFile)
if err != nil {
    return err
}
defer imgFile.Close()

signatureImage, _, err := image.Decode(imgFile)
if err != nil {
    return err
}

opts := annotator.NewSignatureFieldOpts()
opts.Rect = []float64{10, 25, 110, 75}
opts.Image = signatureImage
opts.ImagePosition = annotator.SignatureImageRight
opts.WatermarkImage = watermarkImage

image.Decode needs a decoder registered for the format. UniPDF’s model and core packages already pull in PNG, JPEG and GIF, so those work without any extra import; anything else needs its own blank import, such as _ "golang.org/x/image/webp".

With AutoSize left on, each image is scaled to fit its half of the box, keeping its aspect ratio, and centered there. Turn AutoSize off and the image is drawn at one point per pixel from the lower left corner of its area, which for a scan of any size means it spills well outside the box. Keep AutoSize on unless the image is already sized in points.

Limitations

Images are re-encoded with opts.Encoder, a flate encoder by default. That is lossless and fine for line art, but a photographic JPEG will grow considerably; set opts.Encoder = core.NewDCTEncoder() to keep it JPEG-compressed.

An alpha channel is preserved as a soft mask on the embedded image, so a signature scan with a transparent background shows the box fill through it. The mask is encoded with the same opts.Encoder, which is a reason to leave that at the default when the image has transparency. A fully opaque image has its alpha dropped, so nothing is wasted on images that do not need it.

The watermark covers the full Rect regardless of ImagePosition, and it is drawn after the background fill but before the text and the signature image, so it sits behind both. There is no opacity control for it; FillOpacity applies to the background rectangle only.

Rect is required in practice when an image is involved. Without it the box is sized from the text and then doubled to make room for the image, which is rarely the size you want.

Run the example

The example puts five signature fields on every page, one for each image position, one with the watermark but no signature image, and one with the image alone and no text. All five share a single signature dictionary.

git clone https://github.com/unidoc/unipdf-examples.git
cd unipdf-examples/signatures
go run pdf_image_sign_appearance.go <IN.pdf> <IMAGE> <WATERMARK_IMAGE> <OUT.pdf>

If this is your first time using UniPDF, follow the getting started guide to create an API key and set up your development environment.

Sample output