Presented a New Method in Determining the Amount of Similarity Faces Using Symbols and HSV Colored Spectrum Histograms Hossein Shirgahi 1,a and Najmeh Danesh 2,b+ 1 Jouybar Branch, Islamic Azad University, Jouybar, Iran 2 Sari Branch, Islamic Azad University, Sari, Iran Abstract. A lot of researches have been done about image processing these years. One of the most widely used fields is face recognition based on appearance features. In this research 250 samples of different people’s faces has been taken in RGB mode as the input data and symbols and angles between applied symbols and considered parameters have been achieved and have been saved in the database. Then every image is changed to HSV and its histogram is made. Because of the importance of H spectrum, we considered 18 levels for its H’s histogram and 3 levels for the histograms’ of S and V. then we calculate the probability of histograms’ levels of colored spectrum of gotten images and we register them. Finally, we determine the similarity of images by using symbols features and colored spectrums’ histograms. As experimental results, the performance of this method is 12.5 percent more efficient than the similarity determination method based on symbol. Furthermore, this method is 19 percent more efficient than methods such as similarity determination method based on objective and spatial and similarity determination method which is only based on colored histogram. Keywords: image processing, colored histogram, HSV colored mode, symbolic images. 1. Introduction Image processing is one of the most widely used scopes in computer. This is a part of computerized images’ scopes that these computerized processes are performed based on human’s vision’s system. For these kinds of usages, it is essential to know the quality of human’s vision system. The main contexts of image processing‘s scopes are image retrieval, image improvement and image compression. Image retrieval is a most widely used contexts as analyses on images to reach some scientific, economical and security goals [6]. In most applications which are produced economically to do image retrieval for real environments, there are two important factors: running speed and accuracy of these applications. Also they are important in face recognition of image retrieval too. The most important parameter in produced software and hardware is accuracy, because image retrieval for face recognition is used in military and high secure environments. After accuracy, the speed of decision is important too. The previous works’ accuracy is low because they considered only one or two features of an image. R.R. Venkateswara and et al (2008) presented a new method for images’ texture retrieval based on wavelets multi mode by Markov hidden tree. The features of images texture are extracted by using the effect of textures which are extended of sub groups’ wavelets’ coefficient. Experiments ‘results show that this method reaches more punctual answers rather than previous methods which are based on wavelet in image texture retrieval [7]. + a Hossein.Shirgahi@gmail.com, b Najmeh.Danesh@Gmail.com 2011 International Conference on Computer and Software Modeling IPCSIT vol.14 (2011) © (2011) IACSIT Press, Singapore 78