Showing papers by "Tanuja Sarode published in 2009"

Image retrieval using color-texture features from dct on vq code vectors obtained by kekres fast codebook generation

Abstract: novel technique for image retrieval using the color- texture features extracted from images based on vector quantization with Kekre's fast codebook generation is proposed. This gives better discrimination capability for Content Based Image Retrieval (CBIR). Here the database image is divided into 2x2 pixel windows to obtain 12 color descriptors per window (Red, Green and Blue per pixel) to form a vector. Collection of all such vectors is a training set. Then the Kekre's Fast Codebook Generation (KFCG) is applied on this set to get 16 codevectors. The Discrete Cosine Transform (DCT) is applied on these codevectors by converting them to column vector. This transform vector is used as the image signature (feature vector) for image retrieval. The method takes lesser computations as compared to conventional DCT applied on complete image. The method gives the color-texture features of the image database at reduced feature set size. Proposed method avoids resizing of images which is required for any transform based feature extraction method.

Color-Texture Feature based Image Retrieval using DCT applied on Kekre’s Median Codebook

Abstract: Image retrieval has become imperative area of research because of vide range of applications needing the image data search facility. Most of the research approaches in the area are either database based indexing or image processing based CBIR. The hours need is to combine these parallel going approaches of research to have better image retrieval techniques. The paper proposes a novel technique for image retrieval using the color-texture features extracted from images based on the color indexing using vector quantization. This gives better discrimination capability for CBIR. Here we are dividing the database image into 2x2 pixel windows to obtain 12 color descriptors (Per pixel Red, Green and Blue) per row of window table. Then the Kekre’s Median Codebook Generation (KMCG) is applied on window table to get 256 centre rows. The DCT is applied on this centre row vector to obtain feature set of size 256x12, which is user for image retrieval. The method takes fewer computations as compared to conventional DCT applied on complete image. The method gives the color-texture features of the image database at reduced feature set size.

Fast codebook search algorithm for vector quantization using sorting technique

Abstract: In this paper we propose partial yet efficient codebook search algorithm which uses sorting technique and uses only comparison. Our proposed algorithm does not use Euclidean distance computation and hence it is fastest as compared to other search methods ES, HOSM, DTPC. Form the results it is observed that proposed algorithm gives more MSE as compared to the exhaustive search method but with good execution speed. We also discuss codebook design methods LBG and FCG. The codebooks of different sizes 128, 256, 512 and 1024 are generated using LBG and FCG algorithm. Both the codebook generation algorithms are compared with respect to the execution speed. All the various search algorithms are implemented on the codebooks of different sizes 128, 256, 512 and 1024 obtained from LBG and FCG algorithms. From the results it is observed that FCG codebook gives better performance parameters MSE and PSNR as compared to LBG codebook and among the search algorithm proposed algorithm gives least time to encode the image with slight degradation in image quality.

Fast Codevector Search Algorithm for 3-D Vector Quantized Codebook

Abstract: This paper presents a very simple and efficient algorithm for codebook search, which reduces a great deal of computation as compared to the full codebook search. The algorithm is based on sorting and centroid technique for search. The results table shows the effectiveness of the proposed algorithm in terms of computational complexity. In this paper we also introduce a new performance parameter named as Average fractional change in pixel value as we feel that it gives better understanding of the closeness of the image since it is related to the perception. This new performance parameter takes into consideration the average fractional change in each pixel value. Keywords—Vector Quantization, Data Compression, Encoding, Searching.

Color image segmentation using Kekre's fast codebook generation algorithm based on energy ordering concept

Abstract: In this paper we introduce segmentation approach which uses fast codebook generation algorithm based on energy ordering concept It is specifically designed to segment low-altitude aerial images which can be used as a preprocessing step to 3D reconstruction This approach uses color similarity and volume difference criteria to merge adjacent regions Experiments performed with real aerial images of varied nature demonstrate that this approach does not result in over segmentation or under segmentation allowing large-scale urban scenes to be segmented in an accurate, reliable and fully automatic way The vector quantization seems to give far better results as compared to conventional on-the-fly watershed algorithm

Vector Quantized Codebook Optimization using K-Means

Abstract: In this paper we are proposing K-means algorithm for optimization of codebook. In general K-means is an optimization algorithm but this algorithm takes very long time to converge. We are using existing codebook so that the convergence time for K-means is reduced considerably. For demonstration we have used codebooks obtained from Linde Buzo and Gray (LBG) and Kekre's Fast Codebook Generation (KFCG) algorithms. It is observed that the optimal error obtained from both LBG and KFCG is almost same indicating that there is a unique minima. From the results it is obvious that KFCG codebook takes less number of iterations as compared to LBG codebook. This indicates that KFCG codebook is close to the optimum. This is also indicated by less Mean Squared Error (MSE) for it.

Centroid Based Fast Search Algorithm for Vector Quantization

Abstract: Vector Quantization(VQ) is an efficient technique for data compression and has been successfully used in various applications. In this paper we present new fast codebook search algorithm which uses sorting and centroid technique to search the closest codevector in the codebook. The proposed search algorithm is faster since it reduces number of Euclidean distance computation as compared to Exhaustive search algorithm while keep the image quality imperceptibly close to Exhaustive search algorithm. We have used the mean absolute error as the quality factor since it gives better feel of distortion. Also the proposed algorithm is compared with other codebook search algorithms given in literature and it is found that the performance parameter' average execution time and average number of Euclidean distance computation per image training vector of the proposed algorithm is considerably better compared to most of them.

Bi-Level Vector Quantization Method for Codebook Generation

Abstract: Vector Quantization is lossy data compression technique and has various applications. Key to Vector Quantization is good codebook. Once the codebook size is fixed then for any codebook generation algorithm the MSE reaches a value beyond which it cannot be reduced unless the codebook size is increase. In this paper we are proposing bi-level codebook generation algorithm which reduces mean squared error (MSE) for the same codebook size. For demonstration we have used codebooks obtained from well known Linde Buzo and Gray (LBG) algorithm. The proposed method is general and can be applied to any codebook generation algorithm.