|
|
|
|
||||
|
© 2000 by Michael Cervantes. All Rights Reserved.
|
|
A few days ago, a Corel user was complaining in Corel's Newsgroups because he has purchased Corel Gallery and he wanted to use its photos for high quality printing. He said that those photos have pixels everywhere, and he thought that his money has gone for nothing. Corel's photos saved me a thousand times, and I know that for the money we pay they are an excellent resource. I remember what my dear friend and "maestro" Dan Margulis said: " There are not bad images, there are bad operators."
I want to share this method with you, first to help those in need, and second to prove that Corel's images are not bad. I will be using some very low resolution images to illustrate this method and they may not display as expected in your browser. But if you practice on your system, you will see the results in no time.
Let's start by opening Corel PHOTO-PAINT 9. I searched for an example image in Corel Gallery CDs. I found this image with three beautiful angels in very bad shape.
Note: If you are using Photoshop, the steps are identical. The only difference will be the command required to bring up the dialog boxes.
The first thing to do while correcting an image is to analyze it. What do we have here? A great image with a color cast. Because it is a 72 DPI WI (wavelet) compressed picture, it has bad pixels and artifacts caused by the compression. Editor's Note: The same is true of images compressed with JPEG format.
What is our objective? This image will be printed at 133 LPI (lines per inch) in a magazine. The pressmen allow us a Total Ink Coverage of 280%. Total Ink = C+M+Y+K. This tell us that our darkest shadow will total 280%, when we add the percentage values of cyan, magenta, yellow and black. The final image will be in CMYK color space, 7 inches widths, and our original is in RGB color space, 5.333 inches widths. This tells us that we have to enlarge this picture and increase the resolution. We know how much we will enlarge it, but we don't know the DPI resolution we need to enlarge it correctly. Many of you will say "Yes, I know since it is going to be printed at 133 LPI, we need 266 DPI image." Several people will go to Resample Menu and resample it immediately. Wrong!
It is best to select Generic US Negative Proofing (6) profile in Corel's Color Manager as your separation device. This profile will convert to CMYK with Light GCR (Gray Color Replacement), and 280 Total Ink.
If the image is very critical, not the case with this image, there is one extra step to perform. Rotate the image 45 degrees with antialias selected, then rotate it back. It will push Photo-Paint to perform some calculations and it will start eliminating some of the bad pixels.
Scan Image Calculator tells me that I need 349 DPI to enlarge an image that will be printed at 133 LPI with a quality factor of 2 and our enlargement objective for this image.
Download Scan Image Calculator
Note: You must unzip the downloaded file and create an icon for the EXE file manually.
Select Image | Resample and enter a resolution increase from 72 to 350 DPI. This is not yet the desired enlargement. After Photo-Paint finishes performing calculations, go back to Image | Resample and select enlargement from 5.333 to 7 inches, with Anti-alias and Maintain aspect ratio selected, resolution should have decreased to 266 DPI.
Why did I do it in two steps? You could think it is not the same as going from 5.333 inches 72 DPI to 7 inches 266 DPI in one step. But if we do that, Photo-Paint will have a 72 DPI image to perform the enlargement. This way, Photo-Paint will have 350 DPI to perform it correctly. When Photo-Paint goes to 350 DPI from 72 DPI, it will create new pixels and will destroy bad pixels. This will produce a softer image, that is what we are looking for. Of course we will lost sharpening, but we will resolve this new problem later.
At this point we still have artifacts. Now we go back and fix them. Since the image is in RGB color space, it should be converted to Lab mode. No changes are performed in an image when we convert back and forth between RGB and Lab. Lab is a great independent color space, because it separate brightness and contrast from colors. In Lab we can alter brightness and contrast without affecting colors and vice versa.
Bring up the Channels Docker by selecting Windows | Dockers | Channels. Select the channel "a" in the Channels Docker and apply a Gaussian Blur to eliminate artifacts from magenta and green. Repeat this operation in channel "b" to kill artifacts in yellow and blue. How many pixels do you enter for the Gaussian Blur? The simple answer is to enter a higher number until you don't see any bad pixels in your preview. Don't be afraid. You can blur channel "a" and "b", it won't affect brightness and contrast.
Now in the Lightness Channel, apply Effects | Noise | Remove Moiré to kill what little is there. At this point our opponent is losing the battle. Lab also is the best color space for drastic color changes without affecting brightness and contrast. Since our image has a cast, we are in the right place at the right time to kill it.
Select Image | Adjust | Tone Curve. For those that don't know how to read Lab in Photo-Paint Info Dialog, I suggest you select CMYK as primary color in Info Dialog options. This way you can read color values in CMYK while performing color corrections. Be careful, Lab is a very sensitive space, any small movement in the curve will perform a vast color change. Here are the curves I used:
channel "L"
channel "a"
channel "b"
And here is the result.
Getting better, but not finished. Image is still soft, and there are some more tricks to be performed in CMYK.
Convert the file to CMYK, Image | Mode | CMYK. Reading ink values, I found that some changes should be done using Tone Curve to adjust the image brightness in darker areas. We have to adjust so that it will print fine on a 280% Total ink limit press. Here are the curves I produced.
cyan channel
magenta channel
yellow channel
black channel
The brightest area of the image should now measure C5, M2, Y2, K0, and the shadows should be around C80, M70, Y70, K70. At this point, the image color is ready to print. Now the trick is to sharpen the image and eliminate the remaining bad pixels.
Create two new layers from background. Select the top layer and apply a small amount of Gaussian blur, but enough to kill anything remaining bad pixels. Next select the layer below and apply a drastic Unsharp Mask: Amount 350, 2 pixels, 0 threshold. To preview the effect while applying it, you will need to made the layer above it invisible. In other images these settings may be different, but don't be afraid to apply a big amount of Unsharp Mask because you can regulate it by decreasing layer opacity. You can even apply a level of 500 Unsharp Mask if you please.
Crazy? Don't worry about it. For the third step, select again the top layer (blurred) and choose Object | Clip Mask | Create Clip Mask from Object Transparency. Note that the Clip Mask is selected in Object Docker.
Select the Eraser tool with a setting of 50% Transparency. Soft edges and nib size may vary according to your taste.
Start erasing the areas that you want to be sharp. Eyes, ears, mouth, hair, jewelry, etc. Since the layer below is very sharp, you will see it thru the Clip Mask. It is good to use a Clip Mask because if you make a mistake you can delete it and create a new one without affecting your image.
When finished select Object | Clip Mask | Combine. The Clip Mask will be combined with the object. Now decrease the layer opacity slowly to let some sharpening come up from the layer below in other areas. Do this also to the layer below (sharper) to let some background show through as well. How much your eyes will tell you?
Now, what image is better?
Do you think it will print fine? I am sure of it. More tricks could be performed to make it better, but those will have to be the subject of other articles so stay tuned.
Your opinion is very important to me, please send your comments to mcdesign@tld.net
What do you think about this article?
Did it teach you something new?
Do you think it will help you to improve your images?
Thanks for your time and attention.
Read more articles by Michael Cervantes along with a short bio.
Copyright © 1995–2013 Unleashed Productions, Inc., All Rights Reserved.
