PERBANDINGAN EMPAT ALGORITMA IDENTIFIKASI GAMBAR PORNO; COMPARISON OF FOUR PORN IMAGE IDENTIFICATION ALGORITHMS

Nugroho, Ardhian Heru, Nur Rokhman

2015 | Skripsi | FMIPA

As one of the negative impacts arising from the development of infornation technology today, pornography seems to be the main problem. One type of pornography that develops is pornography in the media image. Therefore, it's necessary for a study that may help in dealing with these problems. One of the parameters that can be used in determining wether an image is pornographic or not is to see wheter the image contains a naked human objects. If an image contains a naked human objects that image should be categorized as porn image (Lin et al, 2003). Morphological Features algorithm, Nudity Detection Algorithm (NDA), Skin Pixel Quantifier (SPQ) algorithm and RSOR algorithm using HSV color model are several algorithms that can be used to detect wheter an image is included in pornographic images or not. The existence of a comparison of several algorithms can assist in determining the good algorithm to classify pornographic image. From the implementation of porngraphic image detection algorithm, Morphological Features Algorithm (with uniform sized image data) produces the highest accuracy values compared to other algorithms. The best results on Morphological Features Algorithm have accuracy of 83.22%, false negatif of 16.00% and false positif of 17.55% by using the YCbCr color model transformation with a resolution of 60x40. In addition it also found that changes in image size, either reduced or enlarged, not overly affect the value of accuracy.

Kata Kunci : Porn Image Detection; Morphological Features; NDA; SPQ; RSOR