|
(Page 2 of 3) Image masking
Let’s have a look at the performance of the program. The main feature of EZ Mask is that it works with the transparency of the foreground image. The plug-in iteratively estimates the transparency value for every pixel in the image, based on a small sample of foreground (what you want to cut out) and background pixels marked by simple strokes on the image and the edges of the masked object are half-transparent and ready for the new background. Thus, ideally, you don’t need any additional smoothing or feathering of the edges either in the plug-in itself or the photo editor. While most selecting plug-ins require just an additional layer for working with the image, in EZ Mask you need a few more preparation steps. You either have an underlying layer with the background or should unlock the foreground image layer first. Then you add a mask to the foreground layer. Finally, you should have the foreground image, and not the mask active when activating the filter. The procedure is well explained in the manual but may need a few iterations to be remembered. And keep in mind that the original image would be changed and make a copy of it before starting the mask. In the plug-in itself you have a choice of two strategies – Stroke or Fill method. The method may somewhat influence the procession time, but not much.According to the manual, the rendering time is based on the size of the image, size of the object being extracted, stroke (slower) vs. filled (faster) trimap, and whether you have a single or multi-processor machine. As for me, I could not notice any considerable difference in procession time. May be the image was too small, though. The processing is quite long and requires a lot of memory in any case.
The difference between the two methods is in the quantity of the pixels referenced to one of the three zones: Foreground, Background, and the Unknown.In the stroke method you make only the outlines of the foreground and background areas, the rest is rendered as unknown. The program then takes into account only the information from the pixels under the strokes. When Fill method is used, the three zones are defined and filled completely. Thus, the plug-in gets more information about the image pixels. Fill method is probably more convenient for more complicated images, with hair, for example. In the image on the right, I was masking the orangutans using the fill mode. The first step was to mark the edge with Unknown Brush. Then I filled the foreground the Background and the image was ready for processing.Once the image is processed, but you see drawbacks, as in the fragment below, you can make corrections in either of the view modes and repeat processing. In the orangutan picture, I only added a few Foreground strokes at the back fur to remove extra transparency.The absence of transparent background view should be mentioned though. It is only accessible when you have no underlying layer as with the orangutan picture. And without it, it is somewhat difficult to correct the processed selection. And the low quality of the preview should be mentioned: you can’t evaluate the quality of the mask sometimes.
Even after pressing Done and returning to PhotoShop you are able to return to EZ Mask and continue from the place you stopped. To do that you should choose ‘Yes’ in answer to ‘Use the previous trimap?’ question. There is a peculiarity in this, however. When previous trimap is used the masked image loses transparency of the edges. It is especially well seen in the Mask View mode where the mask has no grey edges. And the image itself looks strange. That is EZMask has this option but I can hardly see the practical use of it. On simple images, the difference between the masks made with different methods was minimal. | | | | Stroke method is used | Fill method is used |
The key strategy in the fill method is, of course, to be as close to the edge as possible. EZ Mask also works better where the contrast between the foreground image and the background is high. Where the colors are very close and/or the image is not sharp enough, there would be leakages of background color into the hairy image as in the example.Again, the drawbacks are coming forward on a single color background and may not be seen when you insert the cutout into another image. Like with the dog collages below. The dog picture is complex because of fur and the closeness of foreground and background colors on the left side. The girl picture below is complex due to lots of loose hair in the image. Fill method looked to me as the best strategy here as well. And it did work but for some half-transparent areas around the tails, which are concealed by rather dark background. However, EZ Mask did make a good job with hair anyway. You can see it in the fragments below. There are some blotches of background in the first and second fragments. They can be done with, though, by fine ‘tuning’.
The third fragment shows a very good job of the plug-in, lose hair are very well seen there.
|
Reviews 
