Publication: Autenticación de imágenes basada en la extracción de características
Loading...
Official URL
Full text at PDC
Publication Date
2018
Authors
Advisors (or tutors)
Editors
Journal Title
Journal ISSN
Volume Title
Publisher
Citations
Exportar
Abstract
Con el creciente número de aplicaciones de software que permiten alterar imágenes digitales y su facilidad de uso debilitan la credibilidad de una imagen. Este problema unido a la facilidad de distribución de la información a través de Internet (blogs, redes sociales, etc.), ha provocado que la sociedad tienda a aceptar como cierto todo lo que ve sin cuestionar su veracidad. La falsificación de imágenes y el análisis forense de imágenes tiene como objetivo detectar y localizar falsificaciones de imágenes utilizando múltiples pistas que le permitan determinar la veracidad o no de una imagen. Este trabajo propone dos métodos de autenticación de imágenes digitales. El primero mediante el análisis de patrones locales de textura, en este caso, el sistema propuesto combina el patrón binario local con la transformada discreta wavelet y la transformada discreta del coseno para extraer las características de cada uno de los bloques de la imagen investigada. Posteriormente, se utiliza la máquina de soporte vectorial para crear el modelo que permita la verificación de la autenticidad de una imagen. El segundo método propuesto realiza una detección de alteraciones de tipo copy-move dentro de una imagen, utilizando para ello la transfoprmada discreta del coseno. Las características obtenidas de estos coeficientes permite obtener vectores de transferencia, los cuales se agrupan y mediante el uso de un umbral de tolerancia permite determinar si existe o no regiones copiadas y pegadas dentro de la imagen analizada. Los resultados obtenidos de los experimentos llevados a cabo en este trabajo demuestran lo eficaces de los métodos propuestos. Para la evaluación de los métodos propuestos se realizaron experimentos con bases de datos públicas de imágenes falsificadas que son ampliamente utilizadas en la literatura.
With the increasing number of software applications that allow altering digital images and their ease of use, they weaken the credibility of an image. This problem, together with the ease of distributing information through the Ínternet (Blogs, social networks, et.), has caused that, normally, the society accept as true everything that it sees without questioning its veracity. Image counterfeiting has become into a major threat to the credibility of information and forensic image analysis is aimed at detecting and locating image forgeries using multiple clues that allows verification of the authenticity of an image. The second proposed method performs a detection of copy-move alterations within an image, using the discrete cosine transform. The characteristics obtained from these coefficients allow us to obtain transfer vectors, which are grouped together and through the use of a tolerance threshold, it is possible to determine if there are regions copied and pasted within the analyzed image. The results obtained from the experiments carried out in this work, demonstrate the effectiveness of the proposed methods. For the evaluation of the proposed methods, experiments were carried out with public databases of falsified images that are widely used in the literature.
With the increasing number of software applications that allow altering digital images and their ease of use, they weaken the credibility of an image. This problem, together with the ease of distributing information through the Ínternet (Blogs, social networks, et.), has caused that, normally, the society accept as true everything that it sees without questioning its veracity. Image counterfeiting has become into a major threat to the credibility of information and forensic image analysis is aimed at detecting and locating image forgeries using multiple clues that allows verification of the authenticity of an image. The second proposed method performs a detection of copy-move alterations within an image, using the discrete cosine transform. The characteristics obtained from these coefficients allow us to obtain transfer vectors, which are grouped together and through the use of a tolerance threshold, it is possible to determine if there are regions copied and pasted within the analyzed image. The results obtained from the experiments carried out in this work, demonstrate the effectiveness of the proposed methods. For the evaluation of the proposed methods, experiments were carried out with public databases of falsified images that are widely used in the literature.