ISSN (Online) : 2278-1021 ISSN (Print) : 2319-5940 International Journal of Advanced Research in Computer and Communication Engineering Vol. 3, Issue 9, September 2014 Copyright to IJARCCE www.ijarcce.com 7881 Scene Classification Based on Feature Extraction Punith Kumar M B 1 , Dr. P S Puttaswamy 2 , Assistant, Professor, Department, of ECE, BGSIT, Mandya, India 1 Professor, Department, of EEE, PESCE, Mandya, India 2 Abstract: The scene classification plays a vital role in many areas such as video indexing, video compression, video access and others which is the context of the object detection, recognisation and classification. This paper deals with the basic image analysis techniques like RGB, HSV and color histogram. The frame classification can be categorized in to a different classes like, indoor, outdoor, beach, party, roads, river, sky, play ground, etc here we are classifying the frames by considering the TV news video into anchor and reporter frames. Keywords: Scene Classification, RGB, HSV, Color Histogram. I. INTRODUCTION A scene is a basic unit of a story that takes place at a specific location and the time in any one of these changes there will b a new scene. Here we focuses more on the scene identification and its classification to distinguish among the frames in to a group. Hence one can have the clear detail about the content of the frames. The study of scene classification is based on the application and video which are processed will have a slight change in the information carried by the frame as the time changes. This corresponding group of a frame with or without slight changes are called as shot of a given video. There are many ways and techniques are available for the scene classification. This paper considers the color and histogram analysis of a frame. More attention is required during the grouping of frames because the information which represented by the frame so as to keep the variation from person to person and time to time. Emphasizing the fact is that we here an attempt is made to recognize and identify the shot detection in a frames at different scales and not just based on global features or only with local features. It’s expected to be a good work for classifying the frames however it requires the following research work who made on attempt in classification using various technique are represented in the following section with prior knowledge about resources available with frames. Julia Vogel and Bernt Schiele [1] Elaborated techniques of classifying image into One of the main activities in our daily life is eight semantic categories. Torralba A., Oliva A [2] concentrated on lesser properties of frames instead of going for exclusive classifications. We took into the consideration of both the techniques and also referred the similar technique and followed them for some extent. In our work we have employed the combination of two are more techniques to improve the efficiency or at least minimize the complexity involved other techniques. In order to emphasise the importance of the color image analysis for the classification in collaborated with working essentially on gray scale images, scene recognisation means knowing the information about the semantic category and the content of the environment. Basically there are two application of scene recognition has been observed, 1) Object recognisation to decide the class (category) of the scene. 2) Segmentation and processing of objects and try to categorize each scene through its global information estimation. 3) One of the main activity in our life is the ability to distinguish between the things in order to identify them and link them with our prior knowledge this will gives the ability to recognize and interpret the environment around us. Considering the fundamental problem of computer vision i. e enabling computer to see the ways we see the things in the present day we are expecting the machines would have the capability to match with the human vision classification of an object as a table , a ball and scenes etc. II. TYPES In this paper RGB(Red, Green and Blue), HSV(Hue, Saturation and Value) Color models with histogram analysis are used for frame classification. These color models includes the thresholds which are helpful for the classification of scenes. 1) RGB Color Space: The primary components of RGB model are Red, Green and Blue which can be represented on a Cartesian coordinate system as shown in figure 1. In case of the RGB color model there are three primary colors considered are red, green and blue at three are corners and other three are secondary colors and are Cyan, Magenta and Yellow at other three corners followed by black at the origin and white is at the corner farthest from the origin. Point between the black and the white represents the Gray scale level which is represented by dot line, Figure .2 represent the extracted image and RGB color model histogram. Figure 1: RGB Color Model.