If this tool helped you, you can buy us a coffee ☕
Detect whether an image is a photo of a screen using advanced image analysis. Provides confidence scores and preprocessing features.
When you need to verify whether an image is an original or a photo taken of a screen, our recaptured image detection tool analyzes traces of recapture such as screen textures, moiré patterns, light reflections, and geometric distortions. It outputs a classification result (recaptured or original) along with a confidence score. A recaptured image refers to a secondary image created by using a camera or smartphone to photograph a screen, projection, or other display device.
Q: How do you determine if an image is recaptured?
A: Recaptured images typically exhibit screen grids, color deviations, glare spots, or edge distortions. This tool makes a judgment by analyzing these features. For example, a confidence score above 0.9 usually indicates a high probability of it being a recaptured image.
Q: Does preprocessing scaling affect detection accuracy?
A: Yes. Proper scaling standardizes image dimensions and reduces the impact of resolution differences on the model. We recommend using ratio scaling (0.5-0.8) for ultra-high-resolution images, and dimension scaling (while maintaining the aspect ratio) for images with irregular sizes.
Only PNG and JPG formats are supported, and files must be under 5MB. The confidence score is for reference only; critical decisions should be made in conjunction with other evidence. The tool is primarily optimized for detecting electronic screen recaptures and has limited capabilities in identifying photos of printed documents or complex forgery scenarios. Uploaded images are analyzed in real-time and your data will not be stored or shared.
For recaptured document screenshots, we recommend first using dimension scaling to standardize the width to 800px while maintaining the aspect ratio, which can significantly improve the detection rate. A typical example: an image generated by a smartphone photographing a computer screen usually produces moiré patterns (confidence score 0.85+), whereas a direct screenshot will not show such features (confidence score typically below 0.1).
