Tampered image identification method and device

Abstract

The invention aims to provide a tampered image identification method, a model, a training method of the model, computer equipment and a computer readable storage medium. The method comprises the following steps: acquiring an image to be recognized; extracting multiple tampering identification features of the to-be-identified image; combining the multiple tampering identification features into a multi-dimensional image tampering feature; extracting deep image tampering features from the multi-dimensional image tampering features; and inputting the deep image tampering feature into a classifier to obtain a classification result. According to the invention, an image tampering recognition scheme combining prior knowledge and the convolutional neural network is provided, gained tampering recognition features such as surface features and error level analysis features of the image are utilized, and a tampered image recognition model is guided to learn accurate tampering recognition features.

Description

technical field [0001] The invention relates to the technical field of image processing, in particular to a technology for identifying tampered images. Background technique [0002] With the continuous development of image processing technology, people can tamper with images through various algorithms and software, which seriously challenges the authenticity and legality of images. In severe cases, tampering with images may even cause economic and reputational damage to enterprises and individuals. on the impact. [0003] The existing image tampering detection technology is mainly based on the traditional tampering detection method of feature operator. The traditional tampering detection method requires relevant researchers to design feature operators based on manual experience, and select tampering recognition features from a large number of feature operators, which can achieve better results in limited laboratory scenarios, but for a large number of scenes with rich image...

AI Technical Summary

Problems solved by technology

However, conventional computer vision tasks often only need to extract the surface features of the image such as color, texture, shape, and spatial relationship, which cannot be effective for fine-grained features such as image tampering.

For images with complex textures and tampered images with large noise or small tampered areas, the recognition effect is poor, so it is necessary to explore methods for extracting richer image tampering features

Images