How to Increase the Pixel Size of an Image: Pro Workflows

The Mechanics of Upscaling: Interpolation vs. AI Synthesis

When you need to know how to increase the pixel size of an image, you are fundamentally asking software to invent data that does not exist. Traditional upscaling relies on mathematical interpolation—calculating the color values of new pixels based on surrounding existing pixels. While algorithms like Bicubic and Lanczos have improved, they inherently blur edges and amplify noise when pushing past a 150% scale factor.

Modern workflows, however, utilize Convolutional Neural Networks (CNNs) and generative AI to synthesize new pixel data based on training datasets. According to imaging science principles outlined by Cambridge in Colour, traditional interpolation struggles with high-frequency details like text and foliage, whereas AI models can reconstruct these patterns by recognizing structural context.

This guide details exact, platform-specific workflows for increasing pixel dimensions using Adobe Photoshop, Topaz Gigapixel AI, and Affinity Photo V2, including precise parameter settings to avoid the synthetic 'plastic' look common in AI upscaling.

Workflow 1: Adobe Photoshop (Preserve Details 2.0 & Super Resolution)

Photoshop offers two distinct engines for upscaling: the raster-based Image Size dialog and the RAW-based Camera Raw engine. Choosing the right one depends on your source file format.

Method A: Preserve Details 2.0 (For JPEG, PNG, and TIFF)

This is the standard workflow for flattened, compressed images. Preserve Details 2.0 uses a machine-learning model trained specifically to reduce noise while enlarging.

1. Navigate to Image > Image Size (Alt+Ctrl+I / Opt+Cmd+I).
2. Enter your target dimensions. For print, ensure the resolution is set to 300 PPI.
3. Change the Resample dropdown to Preserve Details 2.0.
4. Adjust the Reduce Noise slider. This is the most critical step.

Critical Parameter Warning: Do not leave the Reduce Noise slider at 0%. On compressed JPEGs, a 0% setting will upscale the compression artifacts, resulting in blocky, muddy edges. The empirical sweet spot for standard web JPEGs is between 18% and 25%. For clean TIFFs or PNGs, drop it to 5% to 10% to preserve micro-texture.

Method B: Camera Raw Super Resolution (For RAW and DNG Files)

If you are working with RAW photography, Adobe's Super Resolution engine yields vastly superior results by doubling the linear resolution (effectively quadrupling the megapixel count) while demosaicing the Bayer pattern simultaneously. You can read the exact technical implementation in the official Adobe Camera Raw documentation.

1. Open your RAW file in Camera Raw (or open a JPEG via Filter > Camera Raw Filter).
2. Right-click the image preview and select Enhance.
3. Check the Super Resolution box. Leave 'Raw Details' checked if available.
4. Click Enhance. This generates a new DNG file with the increased pixel dimensions.

Workflow 2: Topaz Gigapixel AI (Standalone & Plugin)

Topaz Gigapixel AI (currently in version 8.x as of 2026) remains the industry standard for aggressive upscaling (4x to 6x) where Photoshop's native tools begin to hallucinate artifacts or blur fine textures. It costs approximately $99 for a perpetual license.

Selecting the Correct AI Model

Gigapixel does not use a one-size-fits-all algorithm. You must select the model that matches your source material's frequency profile:

• Standard: Best for general photography, landscapes, and architecture. Balances edge sharpness with natural texture.
• High Fidelity: Specifically trained on portraits and macro photography. It excels at reconstructing skin pores, iris details, and hair strands without creating the 'waxy' skin effect common in older AI upscalers.
• Art & CG: Optimized for digital paintings, 3D renders, and pixel art. It sharpens hard edges and flat color fields without introducing photographic noise.

Optimizing the Refinement Sliders

Once the AI model processes the image, use the right-hand refinement panel to correct AI-induced anomalies:

• Suppress Noise: Keep this below 20 for film scans or intentional grain. Pushing it to 50+ will strip all micro-contrast, making the image look like a digital painting.
• Remove Blur: Set between 10 and 30. Values above 40 often create 'halo' artifacts around high-contrast edges (e.g., tree branches against a bright sky).
• Recover Original Detail: Blend this slider to 15-20% to reintroduce the original sensor noise and ground the AI-synthesized pixels in reality.

Workflow 3: Affinity Photo V2 (The Budget-Conscious Pro Alternative)

For designers avoiding Adobe's subscription model, Affinity Photo V2 (priced around $70 one-time) provides robust, mathematically precise resampling algorithms. While it lacks the generative AI of Topaz, its Lanczos and Bicubic implementations are highly optimized for upscaling up to 200%.

1. Go to Document > Resize Document.
2. Input your target pixel dimensions.
3. Set the Resample method to Bicubic (Sharper). This is specifically mathematically weighted to maintain edge contrast during enlargement.
4. For images with severe JPEG compression, apply a subtle Unsharp Mask (Filter > Sharpen > Unsharp Mask) *after* resizing. Set Radius to 1.5px, Amount to 80%, and Threshold to 3 levels to target only the upscaled edges without amplifying flat-area noise.

Platform Comparison Matrix

Platform / Engine	Max Recommended Scale	Best Use Case	Pricing Model (2026)
Photoshop (Preserve Details 2.0)	Up to 300%	Compressed JPEGs, Web Graphics	$22.99/mo (Subscription)
Photoshop (Super Resolution)	Exactly 200% (Linear)	RAW Photography, Archival Prints	$22.99/mo (Subscription)
Topaz Gigapixel AI	Up to 600%	Severe Cropping, Low-Res Archival, Portraits	~$99 (Perpetual/Upgrade)
Affinity Photo V2	Up to 200%	Vector-adjacent graphics, Clean PNGs	~$70 (Perpetual)

Edge Cases and Artifact Mitigation

Increasing pixel dimensions often exposes hidden flaws in the source file. Here is how to handle the most common edge cases encountered in professional retouching workflows.

Dealing with Moiré Patterns

When upscaling images of woven fabrics, architectural screens, or fine grids, AI upscalers frequently misinterpret the grid, creating warped, swirling moiré patterns. The Fix: Do not use AI upscaling for these specific regions. Mask out the fabric or grid in Photoshop, upscale the rest of the image using Topaz or Preserve Details 2.0, and use traditional Bicubic Smoother for the masked grid area, followed by a localized Desaturate and Blur to minimize the interference pattern.

The 'Plastic Skin' Syndrome in Portraits

Generative AI models are heavily biased toward smooth skin. When upscaling a low-resolution portrait, the AI will often erase freckles, scars, and natural dermal texture, replacing them with a synthetic, airbrushed surface. The Fix: In Topaz Gigapixel, use the High Fidelity model. Crucially, push the Recover Original Detail slider to at least 30%. In Photoshop, duplicate the original low-res layer, place it above the upscaled layer, set the blend mode to Linear Light, and drop the opacity to 15%. This maps the original sensor noise and skin texture back over the AI-generated base.

Frequently Asked Questions

Does increasing the pixel size change the DPI/PPI?

Pixel dimensions and PPI (Pixels Per Inch) are entirely separate metrics. Increasing the pixel size (e.g., from 1000x1000 to 4000x4000) adds actual data to the file. Changing the PPI from 72 to 300 without resampling merely changes the metadata instruction telling a printer how densely to pack those existing pixels. To increase physical print size without losing quality, you must increase the actual pixel dimensions using the resampling workflows detailed above.

Can I upscale pixel art using these methods?

No. Standard photographic upscalers (Preserve Details 2.0, Bicubic) will blur the hard edges of pixel art. For pixel art, you must use the Nearest Neighbor interpolation method in Photoshop or Affinity Photo, which duplicates exact pixel blocks without anti-aliasing. Alternatively, use the dedicated 'Art & CG' model in Topaz Gigapixel, though manual redrawing is often required for scales above 200%.

Why does my upscaled image look soft on screen but sharp in print?

Monitors display images at a fixed 1:1 pixel mapping (or scaled via OS-level HiDPI). An upscaled 4000px image viewed at 100% zoom on a 1080p monitor requires scrolling, making it difficult to perceive micro-contrast. In print, the 300 PPI density forces the pixels into a physical space where the human eye blends the synthesized edges. Always judge upscaling sharpness by zooming to 100% and evaluating high-frequency edges (like eyelashes or brickwork), not by viewing the 'fit to screen' preview.