HOG Feature Extraction For Image Forgery Detection

Abstract

Image forgery is a prevalent issue in today's digital age. With the increasing availability of sophisticated image editing tools, it has become relatively easy to manipulate digital images, leading to a rise in the number of fake images being circulated online. Image forgery is the process of altering or manipulating an image to mislead the viewers and propagate false information. It is necessary to use forgery detection algorithms to confirm the validity of the digital photographs. The most common method for hiding evidence of image manipulation is to add more noise to the scene. The noise variation in original photos that are not altered is meant to be consistent. The noise no longer spreads evenly throughout the image if the image is fake. This work suggests a technique for detecting forgeries based on hog feature extraction from noise estimations. Once the image has been processed, block segmentation is carried out utilising the Y component of the transformed YIQ picture. Each block of the image is used to extract hog characteristics, and noise is evaluated using PCA. Supervised clustering is a technique that used to group the image's building blocks. The results of the experiments demonstrate that the suggested method detects faked photos more successfully than the earlier method that relied just on noise estimate.