Local Features Matching Assistance-Based Image Forgery Detection

This paper proposed two techniques to find similar regions in a forged image. The first method involves the partitioning of an image into smaller parts and for each block, calculating SIFT keys and descriptor regions. Pyramid level determines block down sampling frequency. A similarity metric, defined as the overall similarity between two SIFT feature families, is used to select the most likely match between two blocks. The second method represents each object/shape by a subset of its border points. Every point is the main point of a little area around it, capturing its exquisite nuances. Images/objects have features that match their patches. Search heuristics like the genetic algorithm (GA) are used to efficiently compare two images or objects based on the overall similarity of two sets of images or form patches. It uses inheritance, mutation, selection, and crossover from natural evolution. Genetic algorithms solve limited-knowledge problems well. They work well in many search settings due to their comprehensive algorithm. According to experiments, our method matches similar regions well. Genetic algorithms (GA) excel in all search spaces and are highly adaptable.