What is the difference between a mask and an alpha channel?

A mask is a grayscale map that defines which areas are visible or affected by an edit. An alpha channel is where transparency is stored inside the image file itself, and a mask is often saved into it when exporting a cutout.

Is inpainting the same as object removal?

Inpainting is the underlying technique of filling a selected area with matching content. Object removal is one of its most common uses, alongside restoring damaged photos and removing text.

What does super-resolution mean?

Super-resolution is the technical term for creating a higher-resolution image from a lower-resolution one using a model that predicts missing detail. In everyday tools it is usually called AI upscaling.

What causes a halo around a cutout?

A halo appears when a thin band of the original background color remains along the edges of the subject. It is most noticeable when the new background contrasts strongly with the old one.

Which formats support transparent backgrounds?

PNG and WEBP both support transparency through an alpha channel. JPG does not, so transparent areas become a solid color when an image is saved as JPG.

AI Photo Editing Glossary: Terms Explained

This AI photo editing glossary explains the terms you are most likely to meet when removing backgrounds, erasing objects and upscaling images. Each entry gives a short, plain-English definition and, where helpful, why it matters in practice. Use it as a quick reference whenever a tool, tutorial or print shop uses an unfamiliar word.

Image basics

Pixel

The smallest unit of a digital image: a single square of color. A photo is a grid of millions of pixels arranged in rows and columns.

Resolution

The number of pixels in an image, written as width × height, such as 3000 × 2000. Resolution decides how much detail an image can hold and how large it can be shown or printed while staying sharp.

Megapixel

One million pixels. A 3000 × 2000 image has 6 megapixels. Useful for comparing cameras, but width and height are more practical when preparing images.

Aspect ratio

The proportion between width and height, such as 1:1 (square), 4:3, 3:2 or 16:9. Matching the aspect ratio of a print or platform avoids unwanted cropping.

PPI and DPI

PPI (pixels per inch) describes how densely an image's pixels are placed when printed; DPI (dots per inch) describes how many ink dots a printer lays down. People often say DPI when they mean PPI. A common guideline for close-up prints is 300 PPI. Our article on DPI vs PPI vs resolution explains the difference in depth.

Color space (sRGB, CMYK)

The range of colors an image can describe. sRGB is the standard for screens and the web. CMYK describes colors made with cyan, magenta, yellow and black inks and is used in many commercial print workflows.

Transparency and cutouts

Alpha channel

An extra layer of information stored alongside the red, green and blue values of each pixel that defines how transparent it is. Fully transparent pixels are invisible, fully opaque pixels are solid, and values in between create soft, partially see-through edges.

Transparent background

An image area with no color at all, so whatever is placed behind the image shows through. It requires a format with an alpha channel, such as PNG or WEBP. See what a transparent PNG is.

Cutout

An image of a subject with its background removed, usually saved as a transparent PNG so it can be placed on any new background.

Mask

A grayscale map that defines which parts of an image are visible or affected by an edit. White typically means fully visible, black fully hidden, and gray partially visible. Background removers produce a mask of the subject; object removers use a mask of the area you brushed over.

Matting

The process of computing precise, soft-edged transparency around difficult details such as hair, fur and semi-transparent fabric. Good matting avoids hard, cut-out-looking edges.

Fringing (halo)

A thin outline of leftover background color around a cutout, often visible when a subject shot on a light background is placed on a dark one, or vice versa.

How AI understands images

Neural network

A type of machine learning model made of many layers of simple calculations. By adjusting millions of internal values during training, it learns to recognize patterns such as edges, shapes and objects.

Training data

The collection of examples a model learns from. A background removal model, for example, learns from many images paired with accurate masks.

Inference

Running a trained model on a new input, such as your uploaded photo, to produce a result. When you use an online AI tool, the processing step is inference.

Segmentation

Dividing an image into meaningful regions. Salient object segmentation, which background removers rely on, finds the main subject and separates it from everything else. For the full process, read how AI background removal works.

Salient object

The part of an image that naturally draws attention, usually the main subject such as a person, product or animal. Automatic background removal tools aim to detect it without user input.

Editing and restoration

Inpainting

Filling in a selected area of an image with new content that matches its surroundings. It is the technology behind object removal: you mark an unwanted element, and the model reconstructs the background behind it. Learn more in what AI inpainting is.

Magic eraser

A common name for an object removal tool that uses inpainting. In Remove Photo Bg, the object remover works this way: brush over the object and the AI fills the area.

Clone stamp and healing

Traditional manual retouching tools found in editors like Photoshop and GIMP. The clone stamp copies pixels from one area to another; healing tools blend the copied texture with surrounding color and light.

Retouching

Any edit that corrects or improves details in a photo, from removing dust and blemishes to straightening lines.

Size and quality

Upscaling

Increasing an image's pixel dimensions. AI upscaling, also called super-resolution, uses a trained model to predict new detail instead of only stretching existing pixels. The Remove Photo Bg image upscaler enlarges images 4×.

Interpolation (resampling)

Traditional mathematical methods for calculating new pixels when resizing, such as nearest neighbor, bilinear and bicubic. Fast and predictable, but they cannot add real detail when enlarging.

Downscaling

Reducing pixel dimensions. Generally safe for quality, and the most effective way to reduce file size for the web.

Pixelation

Visible square blocks caused by showing an image larger than its resolution supports or by heavy compression.

Artifacts

Unwanted visual errors introduced by processing, such as JPEG blocks, color banding in gradients, halos from over-sharpening or smudges from poor inpainting.

Noise

Random grainy speckles in an image, common in photos taken in low light.

Files and formats

Lossy and lossless compression

Lossy compression (JPG, standard WEBP) discards some data to make files much smaller. Lossless compression (PNG, lossless WEBP) keeps every pixel intact.

JPG, PNG and WEBP

The three most common web image formats. JPG suits photos, PNG suits graphics and transparency, and WEBP offers efficient compression with optional transparency. Remove Photo Bg accepts JPG, JPEG, PNG and WEBP uploads up to 16 MB.

Metadata (EXIF)

Information stored inside an image file, such as camera model, settings, date and sometimes GPS location. It can be removed before sharing for privacy.

Put the terms into practice

The fastest way to understand these ideas is to see them in action. Remove a background to create a cutout with an alpha channel using the Remove Photo Bg background remover, try inpainting with the object remover, or watch super-resolution at work in the image upscaler.