Is it possible to make a gradient-transparent/layer masking image using canvas?

I've aded some code here: https://code.google.com/archive/p/canvasimagegradient/ that adds a drawImageGradient function to the CanvasRenderingContext2D. You can draw an image with a linear or radial gradient. It doesn't work in IE, even with excanvas, due to the lack of getImageData/putImageData support.

The following code for example will draw an image with a radial gradient (context retrieve and image load not shown):

var radGrad = ctx.createRadialGradient(
    img.width / 2, img.height / 2, 10, 
    img.width / 2, img.height / 2, img.width/2);
radGrad.addColorStop(0, "transparent");
radGrad.addColorStop(1, "#000");

ctx.drawImageGradient(img, 112.5, 130, radGrad);

The code works as follows:

1. Create a canvas element dynamically and draw the image on it.
2. Retrieve the imageData for this new canvas.
3. Retrieve the imageData for the location on the canvas you want to draw the image on to.
4. Iterate through the destination imageData and update each pixel adding together a percentage (derived from the gradient transparency value) of the image and destination pixel values.
5. Finally put the updated image data back onto the destination canvas.

Obviously performance is an issue as images get larger. The image on https://code.google.com/archive/p/canvasimagegradient/ it takes about 6-10ms to draw. A 1024x768 image takes about 100ms-250ms to draw. Still usable though as long as you're not animating.

To correctly merge two images using a transparency mask it's first necessary to take one of the two images and put it into an off screen canvas, and add the desired transparency mask using context.globalCompositeOperation = destination-out per @Tommyka's answer

var offscreen = document.createElement('canvas'); // detached from DOM
var context = offscreen.getContext('2d');
context.drawImage(image1, 0, 0, image1.width, image1.height);

var gradient = context.createLinearGradient(0, 0, 0, img.height);
gradient.addColorStop(0, "rgba(255, 255, 255, 0.5)");
gradient.addColorStop(1, "rgba(255, 255, 255, 1.0)");
context.globalCompositeOperation = "destination-out";
context.fillStyle = gradient;
context.fillRect(0, 0, image1.width, image1.height);

Then, to actually merge the two images you then need to draw the other image into another canvas, and then simply draw the alpha-composited offscreen canvas on top of that:

var onscreen = document.getElementById('mycanvas');
var context2 = onscreen.getContext('2d');
context2.drawImage(image2, 0, 0, image2.width, image2.height);
context2.drawImage(offscreen, 0, 0, onscreen.width, onscreen.height);

Demo at http://jsfiddle.net/alnitak/rfdjoh31/4/

Its possible to use context.globalCompositeOperation to make the the mask

context.drawImage(img, 0, 0, img.width, img.height, 0,0, img.width, img.height);
context.globalCompositeOperation = "destination-out";
gradient = context.createLinearGradient(0, 0, 0, img.height);
gradient.addColorStop(0, "rgba(255, 255, 255, 0.5)");
gradient.addColorStop(1, "rgba(255, 255, 255, 1.0)");
context.fillStyle = gradient;
context.fillRect(0, 0, img.width, img.height);

This do not do per pixel manipulation and should be faster