Software Source Code Product: Algorithm for Fully Automated Similar Images Search

Visual data processing and usage are highly regulated by governments and companies. In the current world situation, the usage of ready-to-use apps and services can be blocked or forbidden at any time for any reason. Additionally, there is always a risk that third-party applications or services may not cover all the required features. Our software source code product provides the best solution for individuals and organizations who need to search and filter images in a large dataset while minimizing legal and operational risks.

Our algorithm is built to require minimum processing and storage resources, allowing all data and operations to be done locally in a simple and cost-effective environment. By using our software source code product, you can save processing resources and efficiently implement the algorithm in your own code. We provide Python and C code that can be used as is or modified to meet any additional requirements you may have. The complexity of database integration is out of the scope of this product.

The algorithm is based on mathematics and does not rely on machine learning or other resource-intensive techniques. It is particularly useful for systems that work with large amounts of images and need to automatically control and search for similar or duplicate images. It can also be used to identify images in different locations, avoiding sensitive data transmit through non-protected channels, and define search in distributed locations, such as e-commerce platforms, digital image agencies, stock photography websites, and image boards.

Key Features:

• Safe: All images are processed locally, without being sent to any third-party destination. The image content cannot be restored from the resulting metadata, ensuring data privacy and security.


• Fully Controlled: The algorithm parameters can be easily changed to achieve better results and performance, allowing you to meet specific requirements.


• Independent from Image Rotation and Mirroring: The algorithm can accurately match rotated and mirrored images, providing consistent results regardless of the image orientation.


• Independent from Image Scaling: The algorithm handles symmetrical and asymmetrical scaling, till some point, ensuring that images still match even after scaling. Symmetrical scaling has a minimal influence on the image metadata compared to asymmetrical scaling.


• Produces Reusable Metadata: Each image is processed only once, and the resulting metadata can be used to compare images as many times as needed, reducing processing time and resource usage.


• Image File Format Independent: The algorithm works with any common image format, including JPEG, TIFF, BMP, and more. As long as you can open an image file and retrieve its RGB array, our algorithm will work for you.


• Configurable Similarity Threshold: The similarity threshold can be easily configured to adjust the level of similarity required for image matches. The algorithm recalculates the similarity with the new threshold using the same metadata.

Our software source code product includes ready-to-use Python and C code with comments, allowing you to implement the algorithm using different languages and technologies in just a few man-days. It provides a cost-effective and efficient solution for anyone who needs to search and filter images in a large dataset while maintaining full control and management of their data and operations.

Experience the power of our algorithm by checking out the live preview on our website. For an example of the algorithm implementation using a different technology stack, you can explore our SimSearch Demo product for iPhone and iPad on the App Store.