International Journal of Electrical and Electronics Engineering Research (IJEEER) ISSN 2250-155X Vol. 3, Issue 3, Aug 2013, 1-8 © TJPRC Pvt. Ltd. TRAINING-FREE VIDEO INDEXING SCHEME USING LOCAL STEERING KERNEL YOGESH KUMAR & JIMMY GAUTAM ECE Department, Amity University, Noida, Uttar Pradesh, India ABSTRACT This paper is concerned with the detection of salient object in a video sequence. It is assumed that a whenever a new object is to be detected it will appear in the video in a way such that there will be an abrupt change in the frame sequence, which requires us to segment the video in frames so that we could identify the sudden change in the video. A video can be segmented by assuming that there is gradual or abrupt change in the video sequence. After the video is segmented, the abrupt frame is captured and matched using local adaptive kernel regression (LARK) against a single query. KEYWORDS: Video Segmentation, Video Indexing, Object Detection, Regression Analysis, LARK INTRODUCTION When we have large database of images and videos in the form of home videos or videos from news feeds from different channels, they are always abundant in information in the form of space time images. If we want to find a particular object then we have to go through the whole video to search for that object. This search could be based on metadata such as color, keyword, captioning or description to the image, but relying on metadata is not satisfactory since it tend to produce large result with most of them are unsatisfactory only, and it's almost impossible to manually search for key-frame in a video. Content-based image retrieval (CBIR) is the application of computer vision techniques to the image retrieval problem, that is, the problem of searching for digital images in large databases. Content-based indexing means that the search will analyze the actual contents of the image. Most of the image retrieval techniques are based on the some kind of feature extraction or content based algorithm in which resultant feature vector is compared with the feature vector of the query image. The closest image in comparison with the query image from the feature database is returned such that the result images share common elements with the provided query. The term 'content' in this context might refer to colors, shapes, textures, or any other information that can be derived from the image itself. A system that can filter images based on their content would provide better indexing and return more accurate results. Tamura, Mori & Yamawaki (1978) [1] proposed texture based feature in which measures look for visual patterns in images and how they are spatially defined. The relative brightness of pairs of pixels is computed such that degree of contrast, regularity, coarseness and directionality may be estimated. Tushabe and Wilkinson [2] use shape filters to identify given shapes of an image which will often be determined first applying segmentation or edge detection to an image. Recent techniques in image retrieval include boundary detection feature extraction by Dr Kekre H.B, Mishra D, (2011)[3], in the boundary detection method, the binary image is scanned until a boundary is found. Scanning is done on the basis of K- Nearest Neighbor Method. PCA Algorithm for Feature Extraction by Dr Kekre. H.B, Thepade S.D., Maloo A., (2010)[4] a 2-D data is reduced into one-dimensional format by concatenating each row into a long thin vector. The average image i.e the common part of each image in the test data is calculated and subtracted from the original image to get the unique part of the image. Dr Kekre H.B, Mishra D, (2010)[5]