Binary FX: Photoshop Tech Vault 3.0

Binary FX: Photoshop Tech Vault 3.0 | < | >

Reduce Colour Component Artifacts

How To:

Reduce the individual channel artifacts found in full colour images, including JPEG compression damage, halftone de-screening, colour fringing, colour moiré, film grain, scanner CCD noise and digital camera chroma noise.

Colour artifacts are often subtle when the original composite image is viewed, although they can become a major concern once saturation boosting, channel blending/mixing, sharpening and grayscale conversions are attempted. Common colour modes in Photoshop such as RGB and CMYK contain both luminance and hue/saturation variation in each colour channel. Directly removing unwanted channel artifacts can also remove wanted detail. Sadly, this is the method that many people use. A better alternative for dealing with colour component artifacts separates the colour data from the luminance, rather than treating them both as one.

Filtering Colour Modes:

1a. Lab Mode - Lab mode processing for RGB images is often favoured, as one can directly view and independently edit the separated colour component data in the A and B channels from the Lightness data. If you prefer to avoid Photoshop Lab mode, simply use the method outlined below.

1b. RGB/CMYK Mode - In RGB or CMYK modes, the Fade command must be used directly after filtering the composite/master channel, with the Fade command set to colour blend mode. Alternatively, processing a duplicated layer set to colour blend mode and then merging down can often be a more flexible approach as one can process the individual channels independently, although both methods provide the same result of separating colour from tone.

Colour Component Filtering:

2. Combining common noise reduction filters such as Despeckle, Median, Dust & Scratches, Smart Blur, Reduce Noise and Surface Blur at lower settings is recommended. Feathered selections and Motion Blur may be of help for directional colour moiré bands. Gaussian Blur provides a strong averaging affect but also introduces haloing effects around sharp edges at higher settings.

Tip:

As the averaging process of the colour filtering operation can desaturate fine colour detail that is slightly larger than the colour noise and sometimes affect highly saturated hues such as reds, the use of masks or the history brush to restore the original colour component data in these critical areas may be required. Another approach would be to boost these desaturated hues using a hue/saturation adjustment layer set to colour mode (or fading to color mode after the adjustment if not using adjustment layers).

Further Information:

Digital camera images shot in JPEG mode will often require colour averaging to reduce colour noise from higher ISO settings and to clean colour interpolation artifacts found on high contrast edge detail (purple fringing). When shooting in native raw camera format (.nef, .cr2, .dng etc) one may wish to experiment with various raw conversion software's colour/chroma noise reduction and chromatic aberration commands to see how they compare to doing alternative edits in Photoshop. Unlike many digicam images, scans may only require this colour noise reduction in larger low frequency image areas and not on high contrast edges or finer detail. Lineart, text and other sharp edged content in scans will benefit from the use of edge masks to protect this key edge detail from the colour averaging process and any haloing that may result.

Images that are currently in RGB mode can be duplicated and filtered in Lab mode, then blended back over the original image using colour blend mode. This method provides the most flexibility and preservation of luminance data in the original RGB file (one can even choose to change to 16 bpc mode before going to Lab in later versions to further reduce Lab conversion artifacts). If one is not concerned with these issues, one can obviously just convert the original RGB file to Lab mode and filter the A and B channels. A simpler but slightly less effective alternative is to remain in RGB and use colour fades directly after filtering or to use a duped layer set to colour blend mode for filtering.

CMYK images that do not have an accurate ICC profile description should not be converted out of CMYK mode. Even when the correct ICC profile is used to describe the files pixel values, converting a pre-separated image is often a gamble, unless re-separating a duplicate file for a different print condition, such as newsprint. The use of colour blends or fades is recommended while remaining in CMYK mode, so as to preserve the original separation characteristics (black channel, TVI, GCR, etc).

It is commonly advised that one directly filter the Blue or Yellow channel of their photos if they have extreme noise or artifacts. This is not an ideal first step when dealing with noisy colour image channels, as noise and detail are often both removed. Indirect filtering which targets the colour component of an image can reduce artifacts and generate averaged detail, most notably in the Blue or Yellow channel. If artifacts remain after this indirect approach, they can then be addressed directly in the individual channels or in normal blend mode on the composite channel as a secondary measure. Photoshop CS2 also adds the Reduce Noise filter which has the ability to correct individual channels when used in advanced mode (use this filter after indirect filtering). Note: there may be cases where the indirect colour component filtering suggested in this article creates problems for further more aggressive direct luminance filtering.

RGB Channels from a Transparency Drum Scan

(Please wait for Javascript to load each image)

Related Links:

How To: Edge Mask

< | > | ^ | How To Home | Home