Frequently Asked Questions (FAQ)

A Sprite Sheet is an image atlas that combines multiple smaller images into one large image. It is widely used in game development, web animations, and UI design. By consolidating character frames, icons, effects, and other elements into a single image, you can reduce the number of files, decrease loading requests, and improve rendering performance. This tool can automatically identify and split each element in a sprite sheet into individual transparent PNG files.

Canny edge detection is a classic computer vision algorithm proposed by John F. Canny in 1986. It precisely extracts edge information from images through four steps: Gaussian smoothing for noise reduction, gradient direction and intensity calculation, non-maximum suppression, and dual-threshold hysteresis tracking. In this tool, Canny detection is the foundational step for contour extraction — edges are detected first, then connected into complete contours through morphological operations, and finally used to automatically segment individual objects in the image.

The tool provides two Canny threshold parameters: Threshold1 (low threshold) and Threshold2 (high threshold). The low threshold determines the sensitivity for detecting weak edges, while the high threshold sets the standard for identifying strong edges. Start with the default values and adjust gradually: if edges are being missed, lower the thresholds; if too many noisy edges are detected, raise them. Generally, Threshold2 should be set to 2–3 times the value of Threshold1. You can preview the effects of parameter adjustments in real time on the page to find the optimal settings for your image.

Gaussian blur is used to smooth the image and reduce noise interference before edge detection. Sigma controls the blur range (standard deviation) — higher values produce stronger blurring. Amount (kernel size) controls the size of the blur operator — larger values affect a wider pixel range. For noisy images, increasing these parameters helps reduce false edge detection. For detail-rich images where fine contours need to be preserved, use smaller blur parameters.

Morphological operations are image processing techniques used to improve the quality of edges and contours. The Close operation performs dilation followed by erosion, filling small gaps and holes inside contours to connect scattered edges into complete outlines. Dilation expands white regions (edges), connecting adjacent edge fragments together. Increasing the iteration count strengthens the effect, but too many iterations may cause contours of neighboring objects to merge. It's recommended to start with the default values and fine-tune from there.

The Structuring Element defines the range of influence for morphological operations. K1 and K2 control the width and height of the structuring element (in pixels), respectively. Larger structuring elements produce stronger connection effects, suitable for images with severely broken edges. Smaller structuring elements preserve more detail. Typically, K1 and K2 use the same odd value (such as 3×3 or 5×5), but you can also set different aspect ratios based on the characteristics of your image.

Your uploaded images are sent to the server for processing because the contour detection and image segmentation algorithms need to run on the server side. After processing is complete, you can download the results. All uploaded images and processing results are automatically cleared from the server after the associated Token expires. We do not use your images for training models or any other purpose. If you have privacy concerns, please avoid uploading images containing sensitive information. For details, please refer to our Privacy Policy.

Currently, PNG, JPEG/JPG, and WebP format images are supported for upload. The maximum image dimensions are 16,000 pixels wide by 16,000 pixels tall. Each processed image returns a maximum of 1,000 contours (configurable). If your image exceeds these limits, we recommend cropping or scaling it to an appropriate size before uploading.

For the best splitting results, we recommend: 1) Use sprite sheets with high contrast between the background and elements (white, black, or transparent backgrounds work best); 2) Ensure adequate spacing between elements to avoid them being too close together; 3) Start with default parameters — first adjust the Canny thresholds to find approximate contours, then use Close/Dilate operations to connect broken edges; 4) If elements are small or detail-rich, reduce the Gaussian blur parameters to preserve more detail; 5) Use the 'Show contour index' feature to verify that the detected regions are correct.

Transparent backgrounds are the standard output format for sprite sheet splitting. When cropping individual elements from a sprite sheet, the tool preserves only the actual pixel content within the contour bounding box and sets the remaining areas to transparent (Alpha channel = 0). This way, exported PNGs can be directly used in game engines, web design, or other compositing scenarios without needing to remove the background separately. If you need a non-transparent background, you can add a background color using any image editor after downloading.

When you click the 'Download cutouts' button, the tool crops all detected contour regions into individual transparent PNG files and packages them into a ZIP archive for download. Each file is named by sequence number (e.g., 001.png, 002.png, etc.), matching the contour labels shown on the preview canvas. The ZIP file is generated in the browser and does not go through the server — download speed depends on your local processing power and the number of images.

Yes. This tool features a responsive design and works on both phones and tablets. However, since image processing involves significant computation and canvas operations, processing speed may be slower on lower-powered mobile devices. For the best experience when processing large images, we recommend using a desktop browser.

Processing speed primarily depends on: 1) The resolution and total pixel count of the original image — larger images require more computation time; 2) Current server load; 3) Parameter settings — more morphological iteration cycles increase processing time; 4) The number of detected contours — more contours mean longer subsequent cropping and rendering time. Generally, standard-sized sprite sheets (within 1000×1000 pixels) can be processed in a few seconds.

Traditional Photoshop or GIMP slice tools typically require manually drawing slice regions or cutting based on a fixed grid, which is suitable for regularly arranged sprite sheets. RustImage uses contour-detection-based automatic segmentation: it doesn't rely on fixed grids but instead intelligently identifies the actual contour boundaries of each object in the image. This allows it to handle irregularly arranged elements of different sizes, automatically generating the most compact bounding box crop regions. The entire process is fully automatic, requires no software installation, and is completed directly in the browser.

The tool provides two contour bounding box display modes: AABB (Axis-Aligned Bounding Box) is a rectangle parallel to the coordinate axes — it's fast to compute and is used for the final crop export. Min Rect (Minimum Rotated Rectangle / OBB) is the smallest-area rotated rectangle that can enclose the contour, fitting more tightly to tilted object contours. Switching between display modes helps you visually assess the quality of detection results — if the rotated rectangle fits the object well, it indicates that the contour detection is producing good results.