When you get a click, it'll be in window pixel coordinates. You can then translate it into viewport coordinates (which go from 0.0 on the left to 1.0 on the right). You can then translate those into image coordinates. Here's how it would look all together:
viewer.addHandler('canvas-click', function(event) {
var viewportPoint = viewer.viewport.pointFromPixel(event.position);
var imagePoint = viewer.viewport.viewportToImageCoordinates(viewportPoint.x, viewportPoint.y);
console.log(imagePoint.x, imagePoint.y);
});
For more info on the coordinate systems, see: http://openseadragon.github.io/examples/viewport-coordinates/