OpenCV Image and Video Stitcher

Upload multiple images (for list/panorama stitching) OR a single video file (for sequential frame stitching). Video frames are sampled incrementally based on the interval. Use Pre-Cropping to remove unwanted areas before stitching. Adjust other parameters and click 'Stitch'.

Input Gallery (Drag multiple images/video here)

💡 Tip: Hover over the gallery and press Ctrl+V to paste images or video.

If Checked: Preserves alpha channel (RGBA). Unchecked: Treats images as opaque and outputs RGB (Black background for warped areas).

Use Transparency

Crop Top %

Percentage of height to remove from the TOP of each image/frame BEFORE stitching.

0 49

Crop Bottom %

Percentage of height to remove from the BOTTOM of each image/frame BEFORE stitching.

0 49

Stitcher Mode (Image List)

Method for image list stitching. 'SCANS'/'PANORAMA': Use OpenCV's built-in Stitcher (optimized for translation/rotation). 'SCANS': Optimized for images primarily related by translation (like scanning documents or linear camera motion), potentially using simpler geometric models or assumptions internally. 'PANORAMA': Designed for images captured by rotating the camera around a central point. It uses full perspective transformations (Homography) to handle the complex geometric distortions typical in panoramic shots.'DIRECT_PAIRWISE': Skip OpenCV Stitcher and directly use sequential pairwise feature matching (same as video mode or fallback).

SCANS PANORAMA DIRECT_PAIRWISE

Registration Resolution

Scale factor for the image resolution used during feature detection and matching. Lower values (e.g., 0.6) are faster but may miss features in high-res images. 1.0 uses full resolution.

0.1 1

Seam Estimation Resolution

Scale factor for the image resolution used during seam finding (finding the optimal cut line). Lower values (e.g., 0.1) are much faster.

0.05 1

Compositing Resolution

Scale factor for the image resolution used during the final blending stage. -1.0 uses the original source image resolution. Lower values reduce memory usage but might slightly blur the output.

-1 1

Attempts to correct perspective distortions (waviness) common in panoramas. Can increase processing time.

Enable Wave Correction

Exposure Compensation

Method used by the built-in stitcher to adjust brightness/contrast differences between images. 'GAIN_BLOCKS' is generally preferred for varying lighting.

Fixed Base Image Index (0-based)

Specify the index of the image (0, 1, 2...) that should serve as the 'Bottom/Base' layer without blending. Overlapping images will cover it, but brightness will still adjust. Set -1 to disable.

Pairwise Transform Model

Geometric model for pairwise alignment. 'Homography' handles perspective. 'Affine' (Partial/Full) handles translation, rotation, scale, shear (better for scans, less distortion risk). If stitching fails with one model, try another.

Homography Affine_Partial Affine_Full

Blending Method

Algorithm for smoothing seams in overlapping regions when using the detailed stitcher (for video or image list fallback). 'Multi-Band' is often better but slower.

Linear Multi-Band

Adjusts overall brightness difference before blending when using the detailed stitcher. Recommended.

Enable Gain Compensation

ORB Features

Maximum ORB keypoints detected per image/frame. Used by the detailed stitcher (for video or image list fallback).

500 10000

Match Ratio Threshold

Lowe's ratio test threshold for filtering feature matches (lower = stricter). Used by the detailed stitcher (for video or image list fallback).

0.5 0.95

RANSAC Reproj Threshold

Maximum reprojection error (pixels) allowed for a match to be considered an inlier by RANSAC during transformation estimation. Lower values are stricter.

1 10

Max Distance Coeff

Multiplier for image diagonal used to filter initial matches. Limits the pixel distance between matched keypoints (0.5 means half the diagonal).

0.1 2

Max Blending Width

Limits the canvas width during the detailed pairwise blending step to prevent excessive memory usage. Relevant for the detailed stitcher.

Max Blending Height

Limits the canvas height during the detailed pairwise blending step to prevent excessive memory usage. Relevant for the detailed stitcher.

Blend Smooth Kernel Size

Size of Gaussian kernel to smooth blend mask/weights. Must be POSITIVE ODD integer to enable smoothing (e.g., 5, 15, 21). Set to -1 or an even number to disable smoothing.

Multi-Band Blend Levels

Number of pyramid levels for Multi-Band blending. Fewer levels are faster but might have less smooth transitions.

2 7

Stitched Results

Status / Log

Example Configurations

Input Gallery (Drag multiple images/video here)	Use Transparency	Crop Top %	Crop Bottom %	Stitcher Mode (Image List)	Registration Resolution	Seam Estimation Resolution	Compositing Resolution	Enable Wave Correction	Exposure Compensation	Crop Black Borders (Post-Stitch)	Strict No Black Edges (Post-Crop)	Fixed Base Image Index (0-based)	Pairwise Transform Model	Blending Method	Enable Gain Compensation	ORB Features	Match Ratio Threshold	RANSAC Reproj Threshold	Max Distance Coeff	Max Blending Width	Max Blending Height	Blend Smooth Kernel Size	Multi-Band Blend Levels	Sample Interval (ms)	Max Composite Width (Video)	Max Composite Height (Video)