Index color separation is one of the most powerful — and most misunderstood — techniques in screen printing. It solves problems that CMYK halftone process printing cannot: it does away with the angled halftone screens that cause moiré in process work, it works beautifully on textured garments, it handles opaque inks on dark substrates.

You pay for it in screens. Index generally needs more of them than a simulated process separation, sometimes two or three more — because it builds its tone from colors lying side by side rather than from dot size, and opaque inks “mix” with each other poorly. Anyone promising photographic results on six colors is promising too much.

This guide covers what index color separation is, why it works, how to execute it in Photoshop, how to choose the right color count, and a complete troubleshooting section covering the most common problems printers encounter with indexed separations.


What Is Index Color Separation?

Index color separation is the process of reducing a full-color image (which may contain millions of colors) down to a small, defined set of colors — the “index” — and then generating a separate film/screen for each of those colors. Unlike the halftone methods — and that means CMYK process as much as simulated process on opaque inks — index color printing places each ink color side by side, pixel by pixel, with no intentional overlap. In a halftone the tone comes from dots that grow and shrink, and the color from inks overlapping. In index the dot is always the same size, the inks do not overlap, and each one covers its area completely.

Index is a raster too, just a different kind: the dots do not sit in rows at an angle, they are scattered by an algorithm — diffusion dither, which distributes the error of replacing one color with another in a visually random, spatially distributed pattern. The result is a pseudo-random pixel array that, when viewed from a normal distance, blends into smooth gradients and nuanced color transitions.


Index Color vs. CMYK Process: A Direct Comparison

Understanding when to use index color separation versus CMYK requires understanding the fundamental difference in ink behavior.

CMYK Process Printing

CMYK uses four transparent inks (Cyan, Magenta, Yellow, Black) applied in overlapping halftone dots. Color is produced by the eye mixing the overlapping transparent layers. On paper each color gets its own halftone angle; on fabric they all print at the same one. Strengths of CMYK:

Weaknesses of CMYK for screen printing:

Index Color Printing

Index uses a set of opaque spot colors placed in a diffusion dither pattern. No two colors intentionally overlap; each ink sits cleanly on the substrate (or underbase). Strengths of index color:

What is not on that list: a lower mesh count. Index runs on the same mesh counts as a fine halftone. At 180 DPI the dot wants a 120/cm (305/inch) screen — exactly what a 65 LPI halftone wants — and at 160 DPI a 90/cm (230/inch) screen will carry it too. See the mesh table below.

Weaknesses of index color:


Simulated Process — the Method This Comparison Usually Leaves Out

CMYK is not the only alternative to index, and for garment work it is rarely the real one. Simulated process separates the image into a set of opaque spot inks, exactly as index does, but prints each of them as a conventional halftone whose dots grow and shrink. It therefore has the same gamut advantage as index — real inks, so real Pantones and fluorescents — while keeping the tonal behaviour of a halftone.

The difference that matters is where the tone comes from. In simulated process each screen carries a full range from 0 to 100 percent by itself, so a single ink can render a smooth gradient on its own. In index an ink has one value only: it is either on that pixel or it is not, and tone exists solely because neighbouring pixels of different inks sit side by side. Everything else follows from that:

Neither method replaces the other. Simulated process is the better tool for soft, photographic, gradient-led artwork when the press is accurate and the fabric is smooth. Index is the better tool when the image is busy, the fabric is textured, the press has some play in it, or moiré has already cost a run.

There is one more practical difference: a good simulated process separation is mostly judgement — how the channels are built, where they are held back, how much they are allowed to overlap. That is why it is still done by hand, one job at a time. An index separation is a different kind of problem, with far more of it decided by arithmetic, which is exactly why it can be automated.

How Many Colors Does Index Color Need?

This is the most common question and the most important decision in index color separation. The answer depends on two things.

Each additional color adds detail and tonal nuance, particularly in skin tones, skies, and shadow areas.

How many screens you need depends above all on what is in the image – how many steps the tonal transitions in that design have.

And on the color of the garment. On white, the fabric itself is the lightest tone, so white takes no place in the palette and there is no underbase. The same artwork on a dark shirt takes at least one screen more. A greyscale graphic closes in 4–5 screens: underbase, white and two greys.


How to Create Index Color Separation in Photoshop: Step by Step

Photoshop’s Indexed Color mode is the standard tool for index color separation. Here is the complete process.

Step 0: Check the File Before You Touch Anything

Almost every separation that has to be redone from scratch was decided here, before a single button was pressed.

Check that black is black and white is white wherever they are meant to be, .

And where a color is meant to be uniform, check that it really is one color — not a mixture in which 95% of the area holds one value and 5% another, a hair away from it. To the eye it is one flat patch; to the separation it is several colors, which means several screens for one ink.

Zoom in on edges and on flat fields, sample several points of the same field with the eyedropper, and be especially suspicious of files that have been resized or saved as JPEG. Those are what produce this.

Check the color profile by hand: Edit → Assign Profile. That dialog tells the truth about whether the document is untagged or carries a profile. An untagged file should be assigned sRGB, which describes the pixels without changing them.

Step 1: Prepare the Image

Start with an RGB image at whatever resolution it came in. The 160–220 DPI from the mesh table above is the resolution the separation ends at, not the one it starts at (see Step 3b: resampling the file down before separating throws away the smoother result). The image should be:

Do not convert to CMYK before performing index separation — work from RGB.

Step 2: Flatten and Merge All Layers

Image → Flatten Image. All layers must be merged before conversion.

Step 3: Convert to Indexed Color

Image → Mode → Indexed Color In the Indexed Color dialog:

Click OK.

Step 3b: Two Passes — Index High, Then Down

Do not resample the file down to the print resolution before separating. Index it at the resolution it arrives in — 300 DPI, or whatever the file actually is. Then convert back to RGB, resize to the print resolution from the mesh table above, and index a second time on the same palette: Image → Mode → Indexed Color, with Palette set to Previous.

The reason is what the downsampling does in between. Reducing an image that is already indexed squeezes the blocks of flat color together and softens the hard edges between them, and the second pass then works on that softened material. The result is visibly smoother than indexing straight to the low resolution, where every edge is decided once, at full contrast, on a coarse grid.

Reusing the palette in the second pass is not optional. Let Photoshop choose freely and it will pick a different set of colors from the downsampled image — and the palette you are about to review in the next step is then not the palette that ends up on the films.

If the file already arrives at the print resolution there is nothing to do: the second pass has nothing to reduce, and one pass is the whole job.

Step 4: Review and Edit the Color Table

Image → Mode → Color Table This shows the exact index colors Photoshop has selected. Here is where expert judgment comes in: Review each color for printability:

What a palette color actually controls. An entry in the index table decides which pixels go to which screen, and nothing else. Which ink gets mixed in the bucket is the printer’s decision, and the film knows nothing about it. That distinction reverses the usual advice: do not “improve” a palette entry to make it prettier or more printable.

Changing a color in the table only repaints the preview on screen – the film comes out the same, because the same pixels go to the same screen. The stronger shade is taken at the bucket.

A color that is not in the file cannot be added in the palette. It happens that an area reads warmer than its surroundings and the client asks for that warmth on the shirt. Sample a few points of it and the pixels often turn out not to be warm at all — only darker than the field around them, which the eye read as warmth by contrast.

Adding a warm ink to the palette will not help then. That ink lies far from what is in the pixels, so the dither keeps choosing the color that was there before and the area stays as it was. The stronger the warm ink you add, the further away it lies and the fewer pixels it wins. There is no saturation that is both visible and close enough — it is a dead end, not a tuning problem.

The fix is to warm the area in the original with an adjustment layer and extract the colors again. Then the warmth is in the pixels: the palette will find it, the dither will assign it, and nothing undoes that afterwards. Retouch before separation, not in the palette.

Step 5: Check Image Quality

Zoom to 100% and inspect:

If quality is insufficient, go back and increase the color count by 2 and repeat. The difference between 6 and 8 colors is dramatic; between 10 and 12 is subtle.

Step 6: Generate Individual Channel Files

Once the indexed image is approved, each color in the index becomes a separate film. To extract each color: Method A: Convert back to RGB and use channels

  1. Convert Indexed → RGB
  2. Use Select → Color Range to select each index color
  3. Create a layer mask from each selection
  4. Output each layer as a grayscale film

Method B: a plugin written for this method. Two things are worth telling apart here. A halftone RIP — AccuRIP, for instance — does not pull the index colors out into channels. It is an ordinary RIP: it makes halftones out of continuous tone and images the film. It will accept indexed output and image it with a dot, but the separation itself has to be handed to it finished.

IndexColorExtractor is a Photoshop plugin written for this method alone: it extracts the palette, lets you set the opacity and underbase of every ink, and writes the spot channels with registration marks and choke.

Step 7: Film Output

Each color outputs as a grayscale film — black areas indicate where that color’s ink will print. Film should be output at 1200–2400 DPI. Unlike CMYK halftone films, index color films do not require specific halftone angles because the diffusion dither pattern is non-directional.


Choosing the Ink Colors for Index Separation

The quality of an index color print depends not just on how many colors are used, but which colors are chosen. The index palette represents the physical inks that will be mixed and put on press — every color must be achievable in ink.

Working with the Printer’s Ink Set

The ideal workflow is to provide the index color table to your printer and have them match each color to their available mixing system (Pantone, custom lab). For a Dragonfly Colors separation you get the full print specification: every screen in print order, the Pantone number for each one, and the mesh count. Not numbers to convert — the sheet the press works from.


Index Color on Dark Garments: Working with an Underbase

Printing on dark garments with index color typically requires a white underbase. The underbase is a solid or partial coverage white layer that creates a neutral foundation for the index colors to print on. Two approaches to underbase + index color:

Option 1: Full White Underbase, Print All Colors on Top

A full coverage white underbase is flashed, then all index colors print wet-on-wet. This gives maximum color vibrancy but can feel heavy on the garment.

Option 2: Partial Underbase — Let the Garment Be the Shadow

The white does not have to go under the whole design. Leave it out of the dark parts of the image and the garment itself becomes the darkest value in the separation. Less ink on the shirt, a softer hand, and the palette no longer has to carry the shadows — it only has to cover mid-tones and highlights, which is often the difference between needing 10–12 colors and getting a good result with 6–8.

The saving comes from the fact that the same ink gives two shades on one screen: one where it lies on the white of the underbase, another where it goes straight onto the fabric. One screen, two values in the separation.

This is the option worth reaching for on dark garments, and it has nothing to do with choke.

Choke Is Not Part of This — and Index Work Usually Does Not Want It

Choke means printing the underbase a pixel or two smaller than the colors above it, so a misregistered press does not show a white halo around the edge of the print. At 160 DPI one pixel is about 0.16 mm; at 180 DPI about 0.14 mm.

It is an allowance at the outline of a shape, and it belongs to flat graphics — a design with a contour and solid fields, the kind with hard, clean edges. That is where a white fringe would be visible and where taking the underbase in by a pixel costs nothing.

It is not something applied to every dot of a dither. Where a design fades out into diffusion dither there is no edge to protect, and choking simply thins the underbase in the mid-tones — it eats pixels that were carrying the image. On gradients and soft transitions, leave it off.

And it is not what lets the garment show through between colors. That comes from the partial underbase above, from ink opacity, and from the dither pattern itself — not from shrinking the white.


Mesh Selection for Index Color Printing

An index dot is the size of one pixel and never changes — unlike an AM halftone dot, which grows and shrinks. So the resolution you separate at fixes the dot size, and the mesh has to be matched to it.

How fine that dot can go is not decided by whether the screen will hold it. The emulsion will hold a dot smaller than a mesh opening perfectly well. The limit is at exposure and washout: a very small dot on the film is too weak a stencil, light undercuts it, the emulsion hardens partly, and it will not come away when the screen is washed out.

Image resolution Pixel size Mesh for the colors Mesh for the underbase
160 DPI 0.159 mm (159 microns) 120/cm (305/inch) or 90/cm (230/inch) 77–90/cm (196–230/inch)
180 DPI 0.141 mm (141 microns) 120/cm (305/inch) 90/cm (230/inch) or 120/cm (305/inch)
220 DPI 0.115 mm (115 microns) 130–140/cm (330–355/inch) 130–140/cm (330–355/inch)

The underbase has its own column for a reason. It carries the same pixel as the colors, but where possible it goes on a mesh one step coarser: the white layer comes out thicker, whiter and more opaque. On the same mesh as the colors the underbase comes out greyer, and the colors sink into the fabric more and lose saturation.

The coarser mesh comes with one condition: the shop has to be able to wash the pattern’s pixel out of it. If 180 DPI washes out on 90/cm (230/inch), the underbase goes on 90; if not, on 120/cm (305/inch), the same as the colors. At 160 DPI with the colors on 90/cm, the underbase steps down the same way, to about 77/cm (196/inch).

At 220 DPI there is nowhere to step down to. The pixel is so small that it is barely visible on the film, and even on 140/cm it is hard to wash out — the underbase goes on 130–140/cm (330–355/inch), the same as the colors.

Yellow mesh or white. A white thread scatters light under the film and the edge of the dot always gets slightly exposed — the opening in the emulsion comes out smaller and the dot prints finer than it is on the film. A yellow thread does not scatter at all, so for raster work it is simply the safer one. The two also differ in exposure time: white mesh exposes faster, yellow takes longer.

Mind the two mesh scales. Europe counts threads per centimetre, the United States per inch, and the same number means the opposite thing in each: 90/cm is a fine mesh, 90/inch a very coarse one. Both counts are given above so there is nothing to misread.

There is no single floor — the mesh sets the resolution. A screen wants roughly 1.3 to 1.7 mesh openings across one pixel, and that one rule gives a different band for every mesh: about 134–176 DPI on 90/cm (230/inch), about 179–234 DPI on 120/cm (305/inch). 180 DPI is the bottom of the band for a 120/cm screen, the mesh most shops reach for.

160 DPI is the value that fits widest, and the one our plugin defaults to: it is the only resolution that sits comfortably on a 120/cm (305/inch) screen and on a 90/cm (230/inch) screen alike. If a shop prints at 55 lines per inch and goes no finer, 160 is right for it; for more margin on the coarser mesh, drop to 150 or 140. Work below the band for your own mesh and the ink still transfers perfectly well — what you give away is detail, because the pixel itself grows.

220 DPI is the ceiling — and the reason sits in the washout, not on the press. Already at 220 DPI you need a CTS unit printing wax straight onto the emulsion: a film laid against a coated screen is too weak a medium at that dot size. Above 220 DPI you do not go at all — it will not wash out.

Interference. Sometimes moiré appears on the print even though the separation is indexed: the pattern of the pixels has lined up with the pattern of the mesh. There is no rule to it — it is chance.


Troubleshooting Index Color Separation

Problem 1: Visible Pixel Grid / Texture Too Obvious

Symptoms: Print looks “blocky” or “pixel-y” at normal viewing distance; texture is more visible than expected. Causes and fixes:

Problem 2: Flat Shadows, Flat Patches, or a Color That Does Not Belong

Symptoms: shadows with no depth; or a large area that came out as one solid patch of ink where the original had texture; or a color turning up somewhere it has no business being — green in an animal’s fur, for instance.

These look like three separate faults and are usually one: a gap in the tonal ramp of a hue family. It is worth knowing, because the obvious diagnosis — “the palette picked the wrong colors” — sends you in exactly the wrong direction.

The first case: a color that does not belong. Green kept appearing in a large, warm brown area where the original had none. Color matching never chose it there — the green lay far further away than the brown that kept winning. The cause was a gap: between the light brown and the dark one the palette held no intermediate step. Pixels from that range had nothing to be rendered with, the dither spread the resulting error onto their neighbours, and the pushed value eventually crossed a boundary beyond which green was the nearest entry. Adding one brown from the middle of the gap removed it. Five earlier attempts changed which colors were in the palette — what was missing was a step between them.

The second case: a flat patch. Skin tones and a sky came out as one solid ink with no texture at all, although the original had plenty of detail. Nothing foreign came in and nothing mixed — one ink simply served the whole range on its own.

The cure here is not to add a color but to lighten the ink that is already in the palette. It sounds backwards and works like this: the algorithm wants to hit the original, so it has to darken that lightened ink again — and the only way it can darken anything is by mixing in pixels of a darker neighbour. That is exactly where the missing texture comes back.

The ordinary causes are still worth ruling out first:

Problem 3: Skin Tones Look Wrong / Unnatural

Symptoms: Flesh tones appear too orange, too pink, or posterized. Causes and fixes:

Problem 4: Banding in Gradients

Symptoms: Smooth gradients in the original image appear as distinct bands of color in the print. Causes and fixes:

Problem 5: Colors in Print Don’t Match Screen Proof

Symptoms: Printed result looks different from the digital proof on screen, even though the color separation appears correct. Causes and fixes:

Problem 6: Moiré in Index Color Print

Symptoms: Regular repeating pattern visible in the print. Causes and fixes:


Print Order for Index Color Separations

Print order determines how each screen is sequenced on press — and unlike CMYK process printing, where four transparent inks reliably overlay in a fixed order, index color separation places opaque inks side-by-side with only minor edge overlap. Getting the sequence wrong doesn’t just shift color slightly — it can let one ink visually dominate its neighbors in the dither pattern, or cause dot-edge buildup that looks muddy on press.

The General Rule: Dark to Light

As a starting point, sequence screens from darkest to lightest:

  1. White underbase (if printing on a dark or colored garment) always prints first.
  2. Dark, low-opacity colors print next, establishing the shadow structure of the image.
  3. Mid-tones follow.
  4. Light, highly opaque colors (near-whites, pastels, bright yellows) print last, since opaque light inks sit cleanly on top and read as crisp highlights.

This mirrors how a painter builds an image — blocking in darks first, adding highlights last — and it matches how opaque inks physically behave in a wet-on-wet stack: each screen prints on top of the previous one, so a light ink applied late stays vivid instead of being buried under darker ink printed afterward.

Why Dark-to-Light Isn’t the Whole Story

For index separations specifically, two more things come into play:

Hue proximity between neighboring colors. Because index printing places colors side-by-side in a diffusion dither pattern rather than overlaying them, colors that sit close together on the color wheel blend smoothly at their shared edges. Colors far apart in hue (e.g. a saturated red next to a saturated green) create a hard visual seam regardless of print order. When sequencing screens, check which colors actually sit adjacent to each other in the dither pattern — not just their position on a light-to-dark scale.

Some bright, saturated colors need white underneath regardless of “lightness.” A pure yellow or vivid red is technically “light” in tonal value, but both are heavily desaturated by a dark garment unless they sit on white. In practice, these colors often need to print over a dedicated patch of underbase even though the overall sequence still runs roughly dark to light.

Treat the order settled before the run as a starting plan, not a final answer. Once everything is set up on the press, run a test pass and compare the print against the original — that is the only reliable way to see what the sequence is really doing. If something is off, swap two adjacent screens and pull another print. That swap after the first proof is common, even when everything looked correct on screen.


Common Mistakes in Index Color Separation

1. Choosing too few colors and expecting photo quality. A 4-color index can look excellent on clear, simple artwork — the kind built from readable areas of color. For photographic work, plan for at least 8–10 colors. Briefing a client on a 6-color “photo” print on a dark garment will often produce a disappointing result. 2. Relying on the automatically generated palette. The program picks colors statistically — by how much area they cover. That is not the same as a choice that is good for the eye. The automatic palette is a starting point: check that skin tones, brand colors and problematic grays are represented at all, and that no two entries sit so close together that a screen is being wasted. Checking does not mean repainting entries to taste — changing a color in the table does not change the film; the shade is chosen at the bucket. 3. Working at the wrong image resolution. Index color separation at 72 DPI produces large, blocky pixels — visible from across a room. Work at 160–220 DPI at print size. 4. Using the same palette for dark and light garments. A palette optimized for printing on white needs to include dark shadow values. A palette for dark garments should lean toward mid-tones and highlights, using the substrate as the shadow. Using one palette for both will produce either washed-out light garment prints or dark, heavy dark garment prints. 5. Forgetting to check the minimum dot printability. After separation, inspect the films for 1-pixel isolated dots. Isolated pixels in fine highlight areas may not survive exposure and washout if the dot is too small for your equipment. In index work the dot size is not set separately — it follows from the resolution you separate at. The only place to steer it is therefore the resolution and mesh table above: drop the DPI or take a finer mesh. With diffusion dither, isolated single pixels are everywhere by design anyway. 6. Printing index color at the wrong mesh count. The mesh follows the resolution you separated at. At 180 DPI the colors go on 120/cm (305 TPI), and the underbase can stay on 90/cm (230 TPI). At 200–220 DPI the colors move to 130–140/cm (330–355 TPI). The finer the mesh, the thinner the stencil and the smaller the dot it can carry — the table is in the mesh selection section above.


Summary: When to Choose Index Color Separation

Index color separation is the right choice when:

It is not ideal when:

Used correctly, index color separation allows screen printers to achieve genuinely photographic print quality with standard equipment, standard inks, and a fraction of the interference risk.


Dragonfly Colors provides professional index color separations optimized for your specific press, substrate, and ink system. Every separation comes with the full print specification: print order, Pantone numbers and mesh counts. Get in touch to discuss your project.

Photoshop plugin

Doing all of this by hand, on every job?

Every step above – extracting the palette, editing the color table, the underbase, the channel
output – is what IndexColorExtractor does inside Photoshop. It is built for index
separation and nothing else: it works from the measured opacity of your own inks instead of
assumptions, keeps one ink on one screen when that ink needs two appearances, and maps where the
separation lost the original before a single screen is burned.

See IndexColorExtractor