Enlarging digital photographs to several times their original 300 PPI size while maintaining sharp detail is probably the main goal of many interpolation algorithms. Despite this general goal, upscaling results can vary significantly depending on the programs used and the interpolation and sharpening algorithms they implement.

Basics

The problems arise because, unlike film, digital images store images in discrete units: pixels. Any attempt to enlarge the image will correspondingly enlarge these pixels - unless interpolation is applied. Hover over the image on the right to see how even the simplest standard interpolation can improve pixel-induced quadraticity.

Before you dive into this chapter, know that there is no magic wand; best optimization is to start with the highest possible image quality. This means using the right tools: a high-resolution, low-noise camera and a good converter to RAW files. If all of this is present, optimizing your digital photo's enlargement can help you get the most out of your image.

Non-adaptive interpolation overview

Recall that non-adaptive interpolation algorithms always face a trade-off between three defects: jaggies, blur, and boundary halos. The following chart and interactive visual comparison demonstrate each algorithm's place in this three-front war.

The results of scaling performed using the most common algorithms are shown below. Hover over the labels to see how each interpolator performs a given magnification:

*standard interpolation algorithm in Adobe Photoshop CS and CS2

The quality diagram on the right roughly shows the coverage area of ​​each algorithm. The nearest neighbor method is the most susceptible to aliasing, but it and the bilinear method are the least susceptible to boundary halos - they differ only in the different balance between jaggies and blur. You will see how the sharpness of the boundary gradually increases between variations of the bicubic method (3-5), but they are achieved by increasing the gradation and boundary halos. The Lanczos method produces results very similar to bicubic and bicubic sharp in Photoshop, except perhaps with a little more aliasing. They all show some degree of gradation, although aliasing can always be completely eliminated by using image blur (7).

Lanczos and bicubic algorithms are among the most commonly used, probably because they are quite good at choosing between three defects (as is obvious from their location close to the center of the triangle). Nearest neighbor and bilinear methods are not computationally expensive and can therefore be used for augmentation on websites or in portable devices.

Overview of Adaptive Methods

Recall that adaptive algorithms (those that use edge detection) do not treat all pixels equally, but instead adapt to the surrounding content of the image. This flexibility produces much sharper images with fewer artifacts (than would be possible with a non-adaptive method). Unfortunately, they often require longer processing times and are usually more expensive.

Even the most basic non-adaptive methods work quite well at preserving smooth gradients, but they all begin to show their limitations when they try to interpolate close to a sharp edge.

Standard algorithm in Adobe Photoshop CS and CS2
still in research phase, not publicly available

« Genuine fractals"(Genuine Fractals) are probably the most commonly used iterative (or fractal) magnification program. She tries to process the photo in the same way as a file vector graphics- achieving scaling with virtually no loss (by at least in theory). Interestingly, its original purpose was not enlargement at all, it was intended to effectively compress images. Since its introduction, times have changed and disk space is now much more accessible, so it has new uses.

Shortcut PhotoZoom Pro(formerly S-Spline Pro) is another common photo enlarger. When interpolating each pixel, it takes into account the many surrounding pixels and tries to recreate a smooth boundary that passes through all known pixels. To reconstruct the boundaries, it uses a spline algorithm, which is similarly used by car manufacturers when developing new smooth lines for their cars. PhotoZoom has several settings - each designed for a different type of image.

Notice how PhotoZoom produces superior results on the above CG graphics, as it is able to produce a crisp, smooth edge with no jaggies for all the curves in the flag. True fractals introduce fine-scale texture that was not present in the original, and their result is this example no better than bicubic interpolation. It's worth noting, however, that genuine fractals did the best job on the flag tips, while PhotoZoom sometimes breaks them apart. The only interpolator that managed to maintain both smooth, clear boundaries and neat endings is SmartEdge.

Examples from life

The comparisons above showed an increase in theoretical examples, but real-life images are rarely that simple. They have to deal with a range of colors, noise, fine textures and edges that are not so easily distinguishable. The following example contains both fine detail, sharp edges, and a smooth background:

Nearest neighbor method	Bicubic	Bicubic soft	PhotoZoom	Genuine fractals	SmartEdge
With sharpening:	bicubic	bicubic soft	PhotoZoom (standard)	genuine fractals	SmartEdge

All methods except the nearest neighbor method (which simply enlarges the pixels) did an outstanding job considering the relatively small size of the original. Pay special attention to problem areas: in terms of gradation, these are the bridge of the nose, the tips of the ears, the mustache and the belt buckle. As expected, everything worked almost identically when rendering a soft background.

Despite the difficulties that computer graphics caused for real fractals, they literally outdid themselves in this real photo. They created the thinnest mustache, which turned out even thinner than it was in the original image (relative to the others). In addition, they sharply rendered the cat's fur, while avoiding the halo effect along the contour. On the other hand, some may find the resulting fur texture undesirable, so there is also a subjective element in making a decision. Overall, I would say that their result was the best.

PhotoZoom Pro and the bicubic algorithm turned out to be quite similar, except that PhotoZoom produced fewer visible boundary halos and slightly less aliasing. SmartEdge also performed exceptionally well, but is still in development and not available for use. This is the only algorithm that worked well for both computer graphics, and for a real photograph.

Sharpening enlarged photos

Our focus was on the type of interpolation. However, sharpening techniques can have at least an equivalent impact.

Apply sharpening after enlarging your photo to its final size, and not before or during the process. Otherwise, previously invisible halo blur masks will become clearly visible. This effect is similar to that obtained by applying an unsharp mask with a radius larger than the ideal one. Hover over the image on the left (a portion of the previously shown zoom) to see what happens if you apply sharpening before zooming. Notice the increase in the size of the halo around the whiskers and along the contour.

Please also keep in mind that many interpolation algorithms have some sharpening built-in(such as bicubic sharp in Photoshop). It is often impossible to avoid edge exaggeration, since Bayer matrix interpolation itself can also exaggerate edges (and increase visual sharpness).

If your camera does not support RAW format(and you're forced to process JPEGs), make sure the camera's built-in sharpening is turned off or minimized. Enable saving maximum quality JPEG, since compression defects, invisible at the original size, will increase significantly with enlargement and subsequent sharpening.

Because an enlarged photo can become significantly blurrier than the original, upscaled images often benefit more from advanced sharpening techniques. These include convolution reversal, edge hypertrophy fine tuning, multi-radius unsharp mask and new opportunity PhotoShop CS2: Smart Sharpening.

Sharpening and viewing distance

The expected viewing distance of your print may change the depth of field and circle of confusion requirements. Further, the image enlarged for the poster will require a larger blur mask radius than the one shown on the website. The following estimate should not be used as anything other than an approximation; The ideal radius also depends on other factors, such as the subject being depicted and the quality of the interpolation.

The pixel density of a typical display ranges from 70-100 PPI, depending on the resolution setting and screen size. The standard value of 72 PPI when using the above calculator means a mask radius of 0.3 pixels - this is the usual radius that is used for images published on websites. Otherwise, a printing resolution of 300 PPI (standard for photo printers) will produce a mask radius of about 1.2 pixels (also typical).

When Interpolation Becomes Important

Permitting a large billboard on the side of the road never requires so much high resolution, like an art gallery print viewed up close. The following tool displays the minimum PPI and maximum print size that can be used before the eye begins to distinguish individual pixels (without interpolation).

1. Image size. Physical, logical size and resolution
2. Resize the image. Image Size command. The concept of resampling.

Image size. Physical, logical size and resolution.

Image File Size is the physical size of the file in which the image is stored. It is measured in kilobytes (KB), megabytes (MB), or gigabytes (GB). The file size is proportional to the pixel dimensions of the image. How more quantity pixels, the more detailed the image obtained during printing. However, storing them requires more disk space and slows down editing and printing. Thus, when choosing a resolution, a compromise must be made between image quality (which must contain all the necessary data) and file size.

Another factor that affects file size is its format. Due to differences in compression methods used in the formats GIF files, JPEG and PNG, file sizes with the same pixel dimensions can vary greatly. The bit depth of the color and the number of layers and channels also affect the file size.

Photoshop supports maximum dimensions images in pixels equal to 300,000 horizontally and vertically. This limitation determines the maximum permissible size and resolution of the image on the screen and when printing.

About pixel sizes and resolution

The pixel dimensions (image size or height and width) of a bitmap image are a measure of the number of pixels across the image's width and height. Resolution is a measure of the clarity of detail in a raster image and is measured in pixels per inch (ppi). The more pixels per inch, the higher the resolution. In general, a higher resolution image results in a higher quality print.

The same image at 72-ppi and 300-ppi; increased to 200%

The combination of pixel size and resolution determines the amount of image data. If the image has not been resampled, the amount of image data remains the same when the image or resolution is changed individually. When you change a file's resolution, its height and width are changed so that the amount of image data remains the same. The same thing happens when you change the height and width of the file.

Photoshop allows you to define the relationship between image size and resolution in the Image Size dialog box (Image > Image Size). Clear the Interpolation option as there is no need to change the amount of image data. Then change the height, width or resolution of the image. When one of the values ​​changes, the others will be brought into line with the first.

A. Dimensions in pixels are equal to the product of the dimensions of the output document and the resolution.
B. Original dimensions and resolution. Reducing the resolution without changing the pixel dimensions (without resampling).
B. Reducing the resolution while maintaining the same document dimensions leads to an increase in pixel dimensions (resampling)

Resize the image. Resampling.

Changing the pixel dimensions of an image affects not only its size on the screen, but also the quality of the image on screen and when printed, that is, the print size or image resolution.

1. Select Image > Image size.
2. To save the current ratio between height and width in pixels, select Maintain Aspect Ratio. This function automatically changes width when changing height and vice versa.
3. In the Dimension fields, enter values ​​for width and height. To enter values ​​as a percentage of the current dimensions, select percentage as the unit of measure. The new image file size appears at the top of the Image Size dialog box (the old size is in parentheses).
4. Make sure Interpolation is selected and select an interpolation method.
5. If your image has layers with styles applied to it, choose Scale Styles to scale the effect of the styles on the resized image. This feature is only available if Maintain Proportions is selected.
6. When you have finished changing the settings, click OK.

For best results when creating a smaller image, downsample and then apply the Unsharp Mask filter. To create a larger image, rescan the image at a higher resolution.

Resampling changes the amount of image data when changing its pixel dimensions or resolution. When downsampling (reducing the number of pixels), the image loses some information. When resampling (increasing the number of pixels or increasing the resolution), new pixels are added. The interpolation method determines how pixels are removed or added.

Pixel resampling

A. Downsampling

B. No change

B. Resampling (selected pixels are displayed for each set of images)

Keep in mind that resampling may result in reduced image quality. For example, resampling an image to a larger pixel size reduces its detail and sharpness. Applying the Unsharp Mask filter to a resampled image can sharpen details in the image.

You can avoid resampling by scanning or creating images with a sufficiently high resolution. To view the results of resizing in pixels or printing proofs at different resolutions, resample a duplicate of the original file.

Photoshop resamples an image using interpolation techniques, assigning color values ​​to new pixels based on the color values ​​of existing pixels. You can select the method to use in the Image Size dialog box.

In neighboring A fast but less accurate method that follows the pixels of an image. This technique is used in illustrations containing unsmoothed edges to maintain crisp edges and create a file smaller size. However, this method can create jagged edges that become noticeable when you distort or scale the image, or perform many selection operations. Bilinear This method adds new pixels by calculating the average color value of surrounding pixels. It produces results of average quality. Bicubic A slower but more accurate method based on analyzing the color values ​​of surrounding pixels. By using more complex calculations, bicubic interpolation produces smoother color transitions than neighbor interpolation or bilinear interpolation. Bicubic, smoother A good method for image enlargement based on bicubic interpolation, designed specifically to produce smoother results. Bicubic, clearer A good method for reducing image size based on bicubic interpolation with increased sharpness. This method allows you to preserve the details of the resampled image. If Bicubic Sharper interpolation makes some areas of the image too sharp, try using Bicubic Interpolation.

You can specify the default interpolation method to use when resampling image data in Photoshop. Choose Edit > Preferences > General (Windows) or Photoshop > Preferences > General (Mac OS), and then choose a method from the Image Interpolation menu.
In preparation images for printing It is useful to set the image size by specifying the print dimensions and image resolution. These two parameters, called document size, determine the total number of pixels and therefore the file size of the image. The document size also determines the base size of the image when placed in another application. You can control the print size using the Print command, but changes made by the Print command will only affect the printed image—the image file size will not change.
If resampling is used for a given image, you can change the print dimensions and resolution independently of each other (thereby changing the total number of pixels in the image). If resampling is turned off, you can change either the image dimensions or resolution - Photoshop will automatically change the remaining value, maintaining the total number of pixels. Typically, to get highest quality The print must first be resized and resized without resampling. Only then, if necessary, can resampling be performed.

1. Choose Image > Image Size.
2. Change the pixel dimensions, image resolution, or both.
   • To change just the print dimensions, or just the dimension and proportionally change the total number of pixels in the image, choose Interpolation, and then choose an interpolation method.
   • To change the print size and resolution without changing the total number of pixels in the image, do not select Interpolation.
3. To save the current ratio between the height and width of the image, select "Save Aspect Ratios". This function automatically changes the width when the height changes and vice versa.
4. In the Print Size field, enter the new height and width values. If necessary, select a new unit of measurement. Note that the Width field in the Columns feature uses the width and spacing between columns specified in the Units and Rulers settings.
5. Enter a new value in the Resolution field. If necessary, select a new unit of measurement.

To restore the values ​​in the Image Size dialog box to their original values, Alt-click (Windows) or Option-click (Mac OS) the Restore button.

Resizing and rotating the canvas. Canvas Size command.

Rotate or flip the entire image

You can use the Rotate Image commands to rotate or flip the entire image. These commands cannot be applied to individual layers, slices of layers, outlines, or the borders of selections. You can rotate a selection or layer using the Transform or Free Transform commands.
Rotate images
A. Flip the canvas horizontally
B. Original image
B. Rotate the canvas vertically
D. Rotate 90° counterclockwise
D. 180°
E. Rotate 90° clockwise

From the Image menu, choose Image Rotation, then from the submenu, choose one of the following commands.

• 180° — Rotate the image by 180°.
• 90° clockwise — Rotates the image 90° clockwise.
• 90° counterclockwise — Rotates the image 90° counterclockwise.
• Freely—Rotate the image by a specified angle. When you select this option, you must enter an angle between 359.99 and 359.99 degrees in the text box. (In Photoshop, you can set the rotation to clockwise or counterclockwise using the CW or CW options.) Click OK.

Note. Rotating an image is a permanent edit that changes the actual information of the image file. If you want to rotate an image for viewing without making permanent changes, use the Rotate tool.

Changing the Canvas Size

The canvas size is the full editable area of ​​the image. You can use the Canvas Size command to increase or decrease the size of the image canvas. Increasing the canvas size adds space around the existing image. When you reduce the canvas size, the image is cropped. When you increase the canvas size of an image with a transparent background, the added area will be transparent. If the image does not have transparent background, then the color of the added canvas will be determined in various ways.

1. From the Image menu, select Canvas Size.
2. Perform one of the following actions.
   • Enter the canvas dimensions in the Width and Height fields. From the pop-up menus next to the width and height fields, select the units of measurement you want.
   • Select the Relative option and enter an amount to add to or subtract from the current canvas size. Enter a positive number to increase and a negative number to decrease the canvas size by the specified amount.
3. To get an anchor point, click the square that shows the desired location of the existing image on the new canvas.
4. Select an option from the Canvas Extension Color menu.
   • “Basic color” - fills a new canvas with the current primary color
   • “Background” - fills a new canvas with the current background color
   • “White”, “Black” or “Grey” - fills the new canvas with the corresponding color
   • “Others” - select a color for the new canvas from the color palette

     Note. You can also open the color picker by clicking the square to the right of the Canvas Extension Color menu.

   The Canvas Extension Color menu is not available if the image does not have a background.

5. Click OK.

Original canvas and base color canvas added to the right side of the image

Crop the image. Crop tool.

Cropping is cutting off parts of an image for the purpose of focusing or improving composition. You can crop an image using the Frame tool and the Crop command. In addition, you can trim pixels using the “Straighten and Crop” and “Trim” commands.

Using the Frame Tool

Crop an image using the Crop tool

Crop an image using the Crop command

1. The part of the image that you want to save is selected using the selection tool.
2. From the Image menu, select Crop.

Crop an image using the Trim command

Crop using the Trim command removes unwanted elements differently than using the Crop command. You can crop an image by clipping out surrounding transparent pixels or background pixels of a specific color.

1. From the Image menu, select Trimming.
2. In the Trim dialog box, select an option.
   • Selecting the Transparent Pixels-based option removes transparency from the edges of the image and leaves the smallest image made up of opaque pixels.
   • Selecting Top Left Pixel Color removes the area that matches the color of the top left pixel in the image.
   • Selecting the Bottom Right Pixel Color option removes the area whose color matches the color of the bottom right pixel in the image.
   • Select image areas to remove: top, bottom, left or right

Transforming Perspective When Framing

One of the parameters of the Frame tool allows you to transform the perspective of the image. This feature is useful when working with images that contain keystone distortion. Keystone distortion occurs when photographing an object from an angular angle. For example, if a tall building is photographed from ground level, the top of the building will appear narrower than its base.

Perspective Transformation Steps
A. Mark the original crop area B. Align the crop area with the edges of the object C. Extend the crop boundaries D.
The resulting image

The Photoshop graphics editor is called raster because it is designed for creating and processing raster images. In simple terms, raster images consist of dots (raster, pixels), the color code of each of them is stored in the computer memory. Therefore, a digital image is just a set of numbers; it can be seen through any output device: a monitor, printer or other printing device.

When working with digital images, you must understand what the image will be used for. For example, if an image is being prepared for publication on a web page, then the intended output device is a monitor. If in the future you want to store the image as a photograph in a photo album, you must first print it on a photo printer. And to print an image in some glossy magazine, a printing press will be used.

Therefore, to prepare digital images for output to a device, you need to know what resolution should be assigned. To display an image through a monitor, the resolution is most often set to 72 or 96 pixels per inch. For printing on a printer - from 120 to 200 pixels/inch. For high-quality typographic printing - from 250 to 300 pixels/inch.

Resolution and geometric dimensions of the image are interrelated. The higher the resolution, the smaller its size, because the higher the resolution of the intended output device, the smaller its raster dots and the smaller the geometric size of the image.

NOTES

Increasing the image resolution increases the file size, which can reduce your computer's performance when processing the image. Therefore, when choosing a high resolution, you must adhere to the golden mean between image quality and file size.

To change the resolution, use the dialog box Image Size(Image Size), and when the resolution is determined and only the geometric dimensions of the image need to be changed, two tools are used: the Frame(Crop) and dialog box Canvas size(Canvas Size).

In Fig. 2.5 window is presented Image Size(Image Size), the name of which is in the menu Image ++(on Mac OS ++).

Dimensions of the image when displayed on the monitor

The dimensions of the image that it will have when printed on a printer

Rice. 2.5. Dialog window Image Size

Notice the two areas where the dimensions are indicated. The top area can only display dimensions in two units: pixels or percentages. Here are the dimensions of the image when displayed on the monitor. Therefore, this area should be used, for example, to prepare an image for use as a desktop picture or for publication on the Internet.

The lower area displays the dimensions that the image will have when printed on the printer. Therefore, this area should be used to determine and set the dimensions of your future printed photo.

If you need to change the image resolution and its geometric dimensions at the same time, then disable the checkbox before making changes Interpolation(Resample Image). Then, when the image is reduced, its dimensions will be increased, and when the resolution is increased, on the contrary, they will be reduced.

ATTENTION!

Reducing the resolution reduces the number of pixels in the image; these pixels are lost forever and cannot be returned by increasing the resolution! So if you have a high quality image, say 300 ppi, you can downsample it to 180 or 72 ppi, but if the original image is low quality, like 72 ppi, you won't be able to upscale it to letterpress quality.

In Fig. 2.6 shows the window Canvas size(Canvas Size), the name of which is in the menu Image(Image). Can also be caused by a key combination ++(on Mac OS ++).

You just need to click on this square

Rice. 2.6. Dialog window Canvas size

This window is designed to change the geometric dimensions of the image; it does not affect the resolution.

Please note the checkbox Relative(Relative). Don't forget to install it before setting new dimensions. If you want to increase the width or height of the image, you should enter a positive number in these fields. To reduce the size, enter a negative number.

It is also convenient to use the switch Location(Anchor), which indicates the direction of resizing. For example, if you need to reduce the height of the image at the top by 2 cm, then in the field Height(Height) you must enter a value of –2 cm when the checkbox is checked Relative(Relative), and switch Location(Anchor) place in the lower middle position. The switch is set by simply clicking on the required square.

Rice. 4.12. Setting standard block parameters

In field Name Enter the title (name) of the standard block. The block name must be unique within the collection in which it will be stored. Different collections may have blocks with the same names.

In the dropdown list Collection select the collection in which the building block will be stored. The collection you select determines the order in which the block is inserted when creating the document. For blocks inserted directly into the document text, it is better to select a collection Express blocks. In addition, the collection Express blocks displayed as a list in the button menu Express blocks(cm. rice. 4.11).

In the dropdown list Category select a block category. Selecting a category determines the position of the block in the button menu Express blocks in Group Text tabs Insert. You can select one of the existing categories or create a new one. The presence of categories makes it easier to find the desired block in the button menu list Express blocks(cm. rice. 4.11).

In field Description You can enter arbitrary text for a clarifying description of the block being created, which will be displayed as a tooltip when you select a block. The field may not be filled in!

In the dropdown list Options select Paste content on the same page so that the building block is inserted on a separate page. Select Paste content into the same paragraph to prevent the content from becoming part of another paragraph, even if the cursor is in the middle of the paragraph. All other content uses the parameter Paste content only.

After you create new building blocks or change their settings, when you end your session in Word 2010, you will be prompted to save your changes to the building blocks file ( rice. 4.13). Click the button Save.

Enlarge image

Rice. 4.13. Saving changes to a collection of blocks

5.3. Inserting a building block into a document

To insert a standard block into a document you need:

Place the cursor where the block will be inserted.

In the tab Insert in Group Text click the button Express blocks.

To insert a building block from a collection Express blocks find this block in the list ( rice. 4.14) and click on it with the mouse.

Enlarge image

Rice. 4.14. Inserting a building block

To insert a block from an arbitrary collection into a document:

From the button menu Express blocks select a team (see rice. 4.14).

In the dialog box Building Blocks Organizer (rice. 4.15) find the desired block and press the button Insert. For easier searching, you can sort blocks by names, collections, categories, templates, and descriptions. To sort, click on the corresponding button ( for example, click on the button Name to sort blocks by name).

Enlarge image

Rice. 4.15. Selecting a standard block

5.4. Changing a building block

5.4.1. Change content

The contents of a standard block cannot be changed.

You can delete an existing block and then create a new one with the same name.

You can replace an existing block.

Insert a building block.

Make the necessary changes.

Save the building block with the same name and the same parameters.

Yes.

5.4.2. Change settings

You can change the name of the building block, the collection in which it is located, add or change the description, etc.

In the tab Insert in Group Text click the button Express blocks and select a team Building Blocks Organizer(cm. rice. 4.14).

In the dialog box Building Blocks Organizer(cm. rice. 4.15 Change properties.

In the dialog box Changing a building block (rice. 4.16) change the parameters and click the button OK.

enlarge image

Rice. 4.16. Changing standard block parameters

When prompted to override a building block, click Yes.

5.5. Removing a building block

In the tab Insert in Group Text click the button Express blocks and select a team Building Blocks Organizer(cm. rice. 4.14).

In the dialog box Building Blocks Organizer(cm. rice. 4.15) find the desired block and click the button Delete.

When prompted to delete a building block, click Yes.

6. Add a cover page

You can add a specially designed title (first) page to the document.

In the tab Insert in Group Pages click the button Front page and in the list that appears, select one of the proposed options ( rice. 4.17).

enlarge image

Rice. 4.17. Selecting a cover page

The page you add may have pictures and other graphical objects, as well as prompts indicating the information you enter ( rice. 4.18).

enlarge image

Rice. 4.18. Front page

Some fields on the cover page are filled in automatically, with information taken, for example, from the file properties. Some fields must be filled out yourself.

When filling out the fields, you do not have to follow the prompts provided. You can change the contents of automatically filled fields. You cannot leave blank fields; they should be removed.

The contents of the fields can be formatted as plain text. You can add text, tables, and graphic objects to the title page.

To remove the cover page in a tab Insert in Group Pages click the button Front page and in the menu that appears, select the command Delete current cover page(cm. rice. 4.17).

Piecewise polynomial methods of image enlargement (bilinear or even bicubic) that are standardly used in many graphics programs exhibit specific artifacts - they blur the boundaries or form steps on sharp changes in brightness.
If significant scaling is required while maintaining clear boundaries and color areas, then contextual fractal methods of image enlargement are more preferable. Fractal methods are free from the above-described disadvantages and, unlike, for example, the bicubic method (which is considered the best of the local polynomial methods), they convey both smooth transitions and clear details of the enlarged image well.

Why might such a method be needed? For example, you want to make a huge poster from a small photograph. And even if, after enlargement, it looks like a painting by an avant-garde artist, at least it will not become a set of shapeless spots, as with traditional enlargement.

By the way, many photographers use fractal enlargement of scanned old photographs, after which old film frames look much better in sharpness and detail than new high-resolution digital images. The differences in the operation of piecewise polynomial and fractal augmentation algorithms are illustrated in the “Highly Artistic Fractal” sidebar.

Particularly interesting is the application of fractal magnification to technical drawings. For example, if we want to make our own card b O larger scale from the small-scale one we have (say, in order to clarify it ourselves), then, naturally, our main priority when increasing it will be to preserve clear lines, roads and boundaries of areas (forests, fields and reservoirs). In this case, it is better for us to use fractal magnification, perhaps even slightly exaggerated.

And whatever O The higher the magnification we will need, the more noticeable the difference will be in the operation of bicubic and fractal algorithms. Note also that in addition to traditional bilinear or bicubic smoothing algorithms for image enlargement, there are other, more complex ones. For example, there are inexpensive plugins for Adobe Photoshop that implement so-called B- or S-spline smoothing. In addition, they are widely used where it is necessary to significantly enlarge an image for large format printing. Some of these modules include Stair Interpolation Pro (SIPro) from Fred Miranda (http://www.fredmiranda.com), which costs less than $20, or Shortcut PhotoZoom (http://www.shortcutpublishing.com), which costs about $50 Printers may also be interested in them because they provide best quality what they offer standard tools. However, today we will consider only fractal magnification, for which you can use one of the following programs.

ADV Fractal

Programs for working with fractals are usually quite complex and expensive. But there is, for example, a simple and free plugin ADV Fractal (http://serioussoft.narod.ru/adv/frk.htm) for Adobe Photoshop, which implements fractal image compression (compression and decompression), as well as image enlargement. ADV Fractal provides the following capabilities: setting image quality parameters, compression ratio, viewing statistics and the current result in real time, the ability to read and write the resulting system of iterated functions to a file, as well as the ability to edit the resulting fractal image - scaling, editing color and spatial orientation of the image , the number of restoration iterations.

The scope of application of this program, according to the author, can be both scientific (for those involved in fractal compression) and practical: it ensures the implementation of high-quality image enlargement - when scaling towards a significant increase in image size, step structures and blurring of color shades do not appear, which is typical for traditional interpolating algorithms.

Genuine Fractals

In our opinion, the best commercial fractal image enlargement program currently available is, perhaps, Genuine Fractals from onOne Software (http://www.ononesoftware.com/), formerly owned by LizardTech.

This is also a family of plugins for Adobe Photoshop (the company declares support for all versions, including CS3) - Genuine Fractals 5 and Genuine Fractals Print Pro 5. Latest updates bring improved upscaling technology, faster playback, several all-new features and a new user interface. According to the developers, Genuine Fractals 5 now allows you to enlarge digital images by more than 1000% without losing quality.

Multiprocessor support has also been added, pre-setting document size, expanded support for layers, built-in Sharpening and Texture, Film Grain, Photoshop Action controllers, and added support for additional file formats. In addition, Genuine Fractals now supports resizing images in Photoshop files that contain multiple layers, including raw text layers, smart objects, alpha channels, etc. Genuine Fractals costs $5 - $160. Trial version of this plugin can be downloaded at: http://www.ononesoftware.com/download.php?action=download&dl_id=6&type=demo.

For connoisseurs of stretching images to incredible sizes without noticeable deterioration in image quality, Genue Fractals plugins will prove to be an indispensable tool.

Extensis.Pxl.SmartScale

The pxl SmartScale module from Extensis Incorporated (http://www.extensis.com/) is similar to and competes with Genue Fractals.

The name Extensis Incorporated is certainly familiar to many Adobe Photoshop users. Plugins such as Mask Pro, Intellihance, PhotoTools and PhotoFrame are still very popular among computer artists around the world.

The pxl SmartScale plugin is the result of adaptation unique technology scaling and compression of PixelLive images for the needs of mass consumers (the developer of the technology is Celartem, which has recently been owned by Extensis). Thanks to SmartScale, any segment of an image or the entire image can be enlarged to gigantic sizes. For example, the percentage of maximum scaling of the source reaches 1600% of the original. What is typical for such algorithms is that such a high indicator is adequately maintained when visualizing the final image.

By using the proprietary PixelLive technology format to save projects, the resulting images can be compressed and thereby significantly facilitate their subsequent transfer to a printing office or to the customer’s computer. Activation of protection against unauthorized use has also been implemented. Artists who have rights to work with PL content will only need installation free program PixelLive Viewer, available as a standalone application or as a plugin.

The cost of pxl SmartScale declared by the manufacturer - $200 - is by no means small, but the tasks that this module solves are often amazing in the quality of its execution.

Highly artistic fractal In order to illustrate the differences in the operation of piecewise polynomial and fractal image enlargement algorithms, let's take a small picture and try to enlarge it tenfold. If you printed this image at ten times magnification without any anti-aliasing algorithm, the result would be frankly “mosaic” (Fig. 1). When using the bicubic smoothing method (which is considered the best among local polynomial ones) in Adobe Photoshop with the default setting to increase image clarity, the bicubic method partially removes the “mosaic”, but exhibits specific artifacts - it blurs boundaries and forms steps on sharp changes in brightness. In addition, the overall impression of a “littered” picture is created (Fig. 2). Setting it to increase the smoothness of the image only worsens the picture - with such magnification it becomes unacceptably “cloudy”, although some of the artifacts at the boundaries are blurred and there is less “garbage” (Fig. 3). Much best effect gives fractal magnification. If you don't want to select parameters, then the most preferable program is Genuine Fractals 5 (Plug-in for Adobe Photoshop). Even if you set all the parameters to default or use the Genuine Fractals Express module, the enlarged image will look a little stylized, but much clearer along the edges, preserving fine details, smooth color transitions and practically no “garbage”. It is possible that with such a significant increase, not all “steps” will be removed, but even what remains will not hurt the eye (Fig. 4). If you are not bothered by strong “stylization”, then you can try to remove any debris altogether and achieve clear and even boundaries and uniform color shading in all areas. By setting the parameters to highest clarity and contrast, we, of course, will inevitably lose a number of small details, but no one will guess about the former “mosaic” (Fig. 5).