Publications (10)0 Total impact
-
Article: Performance Comparison of Four, Eight & Twelve Walsh Transform Sectors Feature Vectors for Image Retrieval from Image Databases
[show abstract] [hide abstract]
ABSTRACT: This paper presents the idea of using Walsh transform to generate the feature vector for content based image retrieval This paper compares the performance of 4, 8 and 12 sectors of Walsh Transform. The meanof real and imaginary values of Walsh sectors in all three color planes are considered to design the feature vector. The algorithm proposed here is worked over database of 270 images spread over 11 different classes. The Euclidean distance is used as similarity measure. Overall Average precision and recall is calculated for the performance evaluation and comparison of 4,8 and 12 Walsh sectors. The overall average of cross over points of precision and recall is above 50%.International Journal of Engineering Science and Technology. 01/2010; -
Article: Query by Image Content Using Colour Averaging Techniques
[show abstract] [hide abstract]
ABSTRACT: The paper presents six innovative content based image retrieval (CBIR) techniques based on colour averaging. The color averaging methods used here are row mean, column mean, forward diagonal mean, backward diagonal mean, row & column mean and forward & backward diagonal mean. Here the feature vector size per image is greatly reduced by using row, column and diagonal mean, then colour averaging is applied to calculate precision and recall to calculate the performance of the algorithm. Instead of using all pixel data of imageas feature vector for image retrieval, these six feature vectors can be used, resulting into better performance and lower computations.The proposed CBIR techniques are tested on generic image database having 1000 images spread across 11 categories and COIL image atabase having 1080 images spread across 15 categories. For each proposed CBIR technique 75 queries (5 per category) are fired on the generic image database and 55 queries (5 per category) are fired on the COIL image database. To compare the performance of image retrieval techniques average precision and recall are computed of all queries. The results have shown the performance improvement (higher precision and recall values) with proposed methods compared to all pixel data of image at reduced computations resulting in faster retrieval.International Journal of Engineering Science and Technology. 01/2010; -
Article: Augmentation of Colour Averaging Based Image Retrieval Techniques using Even part of Images and Amalgamation of feature vectors
[show abstract] [hide abstract]
ABSTRACT: The theme of the work presented here is augmentation of the colour averaging based image retrieval techniques using even part of images given in [1]. The reflection of the original image is taken across horizontal and vertical directions to get a flip image. The even part of the image is obtained by adding original and flip images. It is clear from [1] that the combination of original image with even part gives better image retrieval than the original alone. On the other hand the combination of original image with odd part gives the worst results. Thus the colour averaging techniques like row & column mean (RCM), forward diagonal mean (FDM) and row, column & forward diagonalmean (RCFDM) are applied on the original image and the even part of image. The colour averages (feature vectors) of original image are considered in combination with the even colour averages to get proposed original+even CBIR techniques which are compared with CBIR methods given in [1]. The proposed content based image retrieval(CBIR) techniques are tested on a generic image database having 1000 images spread across 11 categories. For each proposed CBIR technique 55 queries (5 per category) are fired on the image database. To compare the performance of image retrieval techniques average precision and recall are computed for all the queries. The results have shown improved performance (higher precision and recall values) with the proposed methods compared to the simple original image feature vectors. In the discussed image retrieval methods original with even proves to be better than the original. The combination of row, column & forward diagonal means (RCFDM) gives the highest performance in the discussed three methods of feature vector selection for respective CBIR methods (original, original with even).International Journal of Engineering Science and Technology. 01/2010; -
Article: CBIR Feature Vector Dimension Reduction with Eigenvectors of Covariance Matrix using Row, Column and Diagonal Mean Sequences
[show abstract] [hide abstract]
ABSTRACT: Because of the rising demand from wide range of applications theneed of faster and better image retrieval techniques is growing dayby day. Dimension reduction of CBIR feature vectors has gainedmomentum for swift image retrieval. The paper presents few noveltechniques for image retrieval based on principal componentanalysis (PCA). Here feature vectors are eigenvectors ofcovariance matrix obtained using the row mean, column mean,forward diagonal mean, backward diagonal mean and meancombinations of database images. Instead of taking all pixels ofdatabase images for PCA, proposed CBIR methods use meanvectors, thus dimension of feature vectors used for image retrievalis reduced resulting in faster retrieval. The proposed CBIRtechniques are tested on two different image databases, generalimage database (1000 images spread across 11 categories) andCOIL image database (1080 images spread across 15 objectcategories). For each proposed CBIR technique 55 queries are firedon general image database, 75 queries are fired on COIL imagedatabase and net average precision and recall are computed. Theexperimental results show that proposed CBIR techniques givesthe better performance in terms of higher precision and recallvalues with lesser computational complexity than the conventionalPCA based CBIR using complete image data.International Journal of Computer Applications. 01/2010; -
Article: Performance Comparison of Image Retrieval Techniques using Wavelet Pyramids of Walsh, Haar and Kekre Transforms
[show abstract] [hide abstract]
ABSTRACT: The paper presents performance comparison of Wavelet Pyramid based image retrieval techniques using Walsh, Haar and newly introduced Kekre wavelet transforms. Here content based image retrieval (CBIR) is done using the image feature set extracted from Wavelets applied on the image at various levels of decomposition. Here the image features are extracted by applying Wavelets on gray plane (average of red, green and blue) and color planes (red, green and blue components). The techniques Gray-Wavelets and Color-Wavelets are tested on image database having 11 categories with total 1000 images. Total 55 queries are fired on the database. The results show that precision and recall of Wavelets are better than complete transform based CBIR using Walsh and Haar transform, which proves that Wavelets give better discrimination capability in image retrieval at faster query execution speed. The Walsh and Haar Wavelets level-5 outperforms other Wavelets, because the higher level Wavelets are giving coarse color-texture features while the lower level are representing fine color-texture features which are less useful to differentiate the images in image retrieval. Color- Wavelets based CBIR have greater precision and recall than Gray-Wavelets based CBIR.International Journal of Computer Applications. 01/2010; -
Article: Image Tiling to Improve Performance of Image Retrieval Using Color Averaging Techniques
[show abstract] [hide abstract]
ABSTRACT: The research in content based image retrieval has always beennurtured because of the thirst for better and faster image retrievaltechniques. Reducing the feature vector size for faster imageretrieval and till achieving better performance is herculean task.The paper presents 24 novel image retrieval techniques usingcolor averaging methods on row mean, column mean, forwarddiagonal mean, backward diagonal mean, row & column mean,forward & backward diagonal mean, four tiles, sixteen tiles and64 tiles of image. The proposed CBIR techniques are tested ongeneric image database having 1000 images spread across 11categories and COIL image database having 1080 images spreadacross 15 categories. For each proposed CBIR technique 75queries (5 per category) are fired on the generic image databaseand 55 queries (5 per category) are fired on the COIL imagedatabase. To compare the performance of image retrievaltechniques average precision and recall are computed of allqueries. The results have shown the performance improvement(higher precision and recall values) with proposed methodscompared to all pixel data of image at reduced computationsresulting in faster retrievalInternational Journal of Computer Applications. 01/2010; -
Article: Augmentation of Block Truncation Coding based Image Retrieval by using Even and Odd Images with Sundry Colour Spaces
[show abstract] [hide abstract]
ABSTRACT: The augmentation to block truncation coding (BTC) based image retrieval techniques using Even and Odd images with ten different colour spaces is the theme of work given in the paper. Here the original image is reflected across vertical axis to obtain the flip image, then even and odd images are obtained respectively by addition of original with flip and subtraction of flip from original. TheBTC is applied on original image, even image and odd image to get seven different combinational feature sets for content based image retrieval (CBIR) techniques like original, even, odd, original & even, original & odd, even & odd and original & even & odd. Use of ten sundry colour spaces results into total seventy CBIR methods,For experimentation the generic image database having 1000 images spread across 11 categories is used. For each proposed CBIR technique 55 queries (5 per category) are fired on the generic image database. To compare the performance of image retrieval techniques averageprecision and recall are computed of all queries. The results have shown the performance improvement (higher precision and recall values) with these proposed colour- BTC methods. Instead of using just 6 feature vector in BTC, if we perform the image retrieval using the flipping technique wherein the feature vector is increased to 12 and 18,the performance also increases except in the case of normalized rgb colour space. Image flipping helps to improve the performance in all of luminance-chromaticity colour spaces (YUV, YIQ, LUV, Kekre’s YCgCb, YCbCr) as well as non-luminance based colour spaces (XYZ,HSI,RGB,HSV) in comparison of BTC applied on original image. Also overall YUV colour space proves to be the best in all colour spaces for proposed image flipping techniques. The second best performance is given by Kekre’s YCGCb colour space.International Journal on Computer Science and Engineering. 01/2010; -
Article: Performance Comparision of Image Retrieval using Row Mean of Transformed Column Image
[show abstract] [hide abstract]
ABSTRACT: The paper presents innovative content based image retrieval (CBIR) techniques based on row mean of transformed column image as feature vector. For proposed CBIR techniques three different image transforms like Discrete Cosine Transform (DCT), Walsh Transform and Kekre Transform are considered here. For performance comparison the proposed CBIR techniques are tested on gray version of generic image database of 1000 images spread across 11 categories. For each imageretrieval technique 55 queries (5 per category) were fired on theimage database. Average precision and average recall values forall these queries are computed and used for performance comparison. The proposed CBIR method is considered with DC component as part of feature vector as well as without it. In all three transforms and variation of consideration/ignorance of DC coefficient results into total 6 novel proposed CBIR techniques. These techniques are compared with CBIR using full transformed image as feature vector. The results have shown the performance improvement (higher precision and recall values) with proposed methods compared to full transformed image asfeature vector with great reduction in computational complexity.The negligence of DC component causes performance degradation in proposed techniques. The DCT with consideration of DC component gives best performance among the considered image transforms. The erformance ranking of image transforms in proposed CBIR methods can be given as DCT, Walsh transform and Kekre transform.International Journal on Computer Science and Engineering. 01/2010; -
Article: Image Retrieval using Fractional Coefficients of Transformed Image using DCT and Walsh Transform
[show abstract] [hide abstract]
ABSTRACT: The paper presents innovative content based image retrieval (CBIR) techniques based on feature vectors as fractional coefficients of transformed images using DCT and Walsh transforms. Here the feature vector size per image is greatly reduced by taking fractional coefficients of transformed image. The feature vectors are extracted in fourteen different ways from the transformed image. Along with the first being all the coefficients of transformed image, seven reduced coefficients sets (as 50%, 25%, 12.5%, 6.25%, 3.125%, 1.5625% 0.7813%, 0.39%, 0.195%, 0.097%, 0.048%, 0.024%, 0.012% and 0.06% of complete transformed image) are considered as feature vectors. The two transforms are applied on gray image equivalents and the colour components of images to extract Gray and RGB feature sets espectively. Instead of using all coefficients of transformedimages as feature vector for image retrieval, these fourteen reduced coefficients sets for gray as well as RGB feature vectors are used, resulting into better performance and lower computations. The proposed CBIR techniques are implemented on a database having 1000 images spread across 11 categories. For each proposed CBIR technique 55 queries (5 per category) are fired on the database and net averageprecision and recall are computed for all feature sets per transform. The results have shown the performance improvement (higher precision and recall values) with fractional coefficients compared to complete transform of image at reduced computations resulting in faster retrieval. Finally Walsh transform surpasses DCT transforms in performance with highest precision and recall values for fractional coefficients and minimum number of computations up to 0.097% and then DCT takes over.International Journal of Engineering Science and Technology. 01/2010; -
Article: Tumor Demarcation in Mammography Images using LBG on Probability Image
[show abstract] [hide abstract]
ABSTRACT: The ability to improve diagnostic information from medicalimages can be enhanced by designing computer processingalgorithms that is why we proposed new algorithm to detectcancer in mammogram breast cancer images. In this paper weproposed segmentation using vector quantization technique. Herewe used Linde Buzo and Gray (LBG)for segmentation ofmammographic images on probability image. Initiallyprobability of input image is calculated and displayed as a result.In second step a codebook of size 128 was generated forprobability image. These code vectors were further reclusteredin 8 clusters using same LBG algorithm. These 8 images weredisplayed as a result. This approach does not leads to oversegmentation or under segmentation. For the comparison purposewe displayed results of GLCM and watershed segmentationalong with this method.International Journal of Computer Applications. 01/2010;
