Adaptive Image Enhancement

If everything else in this pack feels like too many knobs, this is the answer - a single node that looks at your image, decides what kind of image it probably is, and applies a reasonable combination of noise reduction, contrast, sharpening, and color correction on its own. It's the "I don't want to think about it" node, and it's genuinely useful as a first pass before you decide whether you need one of the pack's more specialized restorers.

How it works

auto_detect_type drives the whole thing: the node classifies the image into one of a handful of buckets - photograph, document (think scanned text or line art), artwork, or a mixed category - and applies a different enhancement recipe depending on which bucket it lands in. A document doesn't want the same contrast curve as a photograph, and definitely doesn't want the same noise reduction (documents want crisp edges preserved, photographs want the noise gone even if it softens fine texture slightly). You can skip the guessing and set image_type manually if you already know what you're feeding it. Each individual technique - noise reduction, contrast, sharpening, color correction - is independently toggleable, so you can turn this into "auto contrast only" or any other subset if the full combo isn't what you want.

The inputs and outputs that matter

• image - required.
• enhancement_level - conservative, moderate (default), or aggressive, controlling overall strength across every technique it applies.
• auto_detect_type (default true) - whether it classifies the image itself.
• image_type (optional, default auto) - override: photograph, document, artwork, or mixed if you want to force the category rather than let the node decide.
• noise_reduction, contrast_enhancement, sharpening, color_correction (all optional, all default true) - toggle any of the four techniques off individually.
• show_analysis (optional, default false) - prints the node's read of the image to console, useful if you want to see what it decided before trusting auto.
• Output: enhanced_image - note there's no processing_info string output on this one, unlike most of the pack's other enhancement nodes; use show_analysis if you want visibility into what happened.

How to install it

Search Eric_Image_Processing_Nodes in ComfyUI Manager, or manually:

cd ComfyUI/custom_nodes
git clone https://github.com/EricRollei/Eric_Image_Processing_Nodes
cd Eric_Image_Processing_Nodes
pip install -r requirements.txt

Restart ComfyUI. This is a classical-CV composite node, not a pretrained model, so it needs only the pack's core requirements (numpy, opencv-python, scikit-image, scipy, PyWavelets) and nothing downloads on first run.

Common issues & troubleshooting

It classified your image wrong. Auto-detection on a genuinely mixed source (a photo of a printed document, say) is exactly the case it's going to struggle with - that's what the mixed category exists for, but the node still has to guess when to use it. If the result looks off, set image_type explicitly instead of trusting auto_detect_type.

Result looks over-processed. Drop enhancement_level to conservative, or turn off individual techniques you don't need rather than accepting the full combo at aggressive.

You want to know what it actually did. Turn on show_analysis and check your console - there's no image-embedded report on this node the way there is on most others in the pack, so this is the only visibility you get.

One specific technique is causing the problem (e.g. color correction shifting hues you don't want). Toggle that one off (color_correction: false) rather than dropping the whole enhancement_level - the four techniques are independent switches for exactly this reason.