Image Similarity Detection (PDQ Algorithm)

An online image similarity detection tool based on the PDQ algorithm commonly used across major platforms. Supports batch comparison analysis to quickly identify duplicate and similar images.

Bug report

Browser execution mode: Your data is processed in your browser and is not uploaded to the server.

Speed and Stability: Processing speed depends on your device and browser. For large batch work, the desktop version may be more stable.

Loading tool, please wait...

If the online tool fails to load or run, try the desktop tool.https://tools.yikeaigc.com/

Tool Usage

Return to old version

1Select original image

Select an original image as the comparison baseline. Supports JPG, PNG, BMP, GIF, and more.

2Select comparison images

Select one or more images to compare with the original. Batch selection supported.

3Detection settings

Similarity rules

≤31- Duplicate image

32-64- Similar image

>64- Dissimilar image

Hash algorithm

PDQ hash algorithm- High-accuracy image similarity detection algorithm open-sourced by Facebook

Show details

Sort by similarity

Instructions

Software Usage Instructions

Select Original Image: Click the "Select Original Image" button and choose the image file to use as the baseline. Common image formats such as JPG, PNG, BMP, and GIF are supported.
Select Comparison Images: Click the "Select Comparison Images" button to choose one or more images to compare with the original image for similarity. Batch selection of multiple image files is supported.
Start Detection: After uploading the images, click the "Start Detection" button. The tool will automatically use the PDQ algorithm to calculate the image hash values.
View Results: After detection is complete, the system will display the Hamming distance value between each comparison image and the original image, and automatically determine the similarity level according to fixed rules:
- Duplicate (Hamming distance ≤31): The image content is basically the same
- Similar (Hamming distance 32-64): The image content is similar but has differences
- Not Similar (Hamming distance >64): The image content differs significantly
Download Results: You can export the detection results as an Excel file, including detailed information such as file name, hash value, Hamming distance, and similarity judgment.

FAQ

A: The PDQ algorithm is a professional image hashing algorithm developed by Facebook. Compared with traditional aHash and pHash algorithms, it offers higher accuracy and robustness. The PDQ algorithm can better handle changes such as image rotation, scaling, and compression, providing more accurate similarity detection results while maintaining high efficiency.

A: The Hamming distance judgment criteria are based on extensive experimental data and industry experience: ≤31 is duplicate, meaning the image content is basically the same; 32-64 is similar, meaning the images have a certain degree of similarity but also differences; >64 is not similar, meaning the image content differs significantly. This standard can effectively distinguish different levels of image similarity in practical applications.

A: The tool supports mainstream image formats such as JPG, JPEG, PNG, BMP, and GIF. It is recommended that each image file be no larger than 10MB. For batch detection, it is recommended to upload no more than 50 images at a time to ensure detection efficiency and stability.

A: No. All image processing is completed locally on your device, and image files will not be uploaded to the server, ensuring your privacy and data security. The tool only calculates image hash values and Hamming distances locally, protecting your image content from being leaked.

A: The tool uses the industry-recognized PDQ algorithm, which offers high accuracy in image similarity detection. At the same time, the Hamming distance calculation method is scientific and reliable, and can accurately reflect the degree of difference between images. It is recommended to confirm detection results based on your actual needs and manual judgment when using it.

A: The tool is suitable for multiple scenarios: image deduplication to clean up duplicate files and save storage space; copyright protection to detect whether images have been stolen; content moderation to identify similar or duplicate uploaded content; image management to organize and classify large numbers of image files; quality control to check similarity changes before and after image processing.

Answer: The PDQ algorithm primarily makes judgments based on the overall structure and color distribution of an image, so it may classify images with similar compositions and similar color tones as similar. This is a characteristic of hash algorithms: they focus more on the image’s overall features than on detailed differences. For more accurate judgment, we recommend combining it with manual review or using multiple algorithms for comprehensive evaluation.