In this paper we present an effective clustering algorithm to generate codebook for vector quantization (VQ). Constant error is added every time to split the clusters in LBG, resulting in formation of cluster in one direction which is 135 0 in 2dimensional case. Because of this reason clustering is inefficient resulting in high MSE in LBG. To overcome this drawback of LBG proportionate error is added to change the cluster orientation in KPE. Though the cluster orientation in KPE is changed, its variation is limited to ± 45 0 over 135 0. KEVR introduces new orientation every time to split the clusters. But in KEVR the error vector sequence is the binary representation of numbers, so the cluster orientation change slowly in every iteration. To overcome this drawback we propose the technique which uses Walsh sequence to rotate the error vector. The proposed technique (Kekre's error vector rotation using Walsh-KEVRW) is based on KEVR algorithm. The proposed methodology is tested on different training images for code books of sizes 128, 256, 512, 1024. Our result shows that KEVRW gives less MSE and high PSNR compared to LBG, KPE and KEVR.

In supervised classification of image database, feature vectors of images with known classes, are used for training purpose. Feature vectors are extracted in such a way that it will represent maximum information in minimum elements. Accuracy of classification highly depends on the content of training feature vectors and number of training feature vectors. If the number of training images increases then the performance of classification also improves. But it also leads to more storage space and computation time. The main aim of this research is to reduce the number of feature vectors in an effective way so as to reduce memory space required and computation time as well as to increase an accuracy. This paper proposes three major steps for automatic classification of image database. First step is the generation of feature vector of an image using column transform, row mean vector and fusion method. Then vector Quantization (code book size 4,8 and 16) is applied to reduce the number of training feature vectors per class and generate an effective and compact representation of them. Finally nearest neighbor classification algorithm is used as a classifier. The experiments are conducted on augmented Wang database. The results for various transforms, different similarity measures, varying sizes of feature vector, three code book sizes and different number of training images, are analyzed and compared. Results show that the proposed method increases accuracy in most of the cases.

The paper presents a new approach of finding nearest neighbor in image classification algorithm by proposing efficient method for similarity measure. Generally in supervised classification, after finding the feature vectors of training images and testing images, nearest neighbor classifier does the classification job. This classifier uses different distance measures such as Euclidean distance, Manhattan distance etc. to find the nearest training feature vector. This paper proposes to use Mean Squared Error (MSE) to find the nearness between two images. Initially Independent Principal Component Analysis (PCA),which we discussed in our earlier work, is applied to images of each class to generate Eigen coordinate system for that class. Then for the given test image, a set of feature vectors is generated. New images are reconstructed using each Eigen coordinate system and the corresponding test feature vector. Lowest MSE between the given test image and new reconstructed image indicates the corresponding class for that image. The experiments are conducted on COIL-100 database. The performance is also compared with  distance based nearest neighbor classifier. Results show that the proposed method achieves high accuracy even for small size of training set.

Evolution of medical imaging has turned out as a boon for medical industry as it provides efficient diagnosis and monitoring of diseases. Compression of medical images helps accommodation of large medical data in limited storage space and fast transmission. The main aim of this paper is to compress medical images with no loss of clinical data using a lossless and adaptive prediction technique. The paper presents a prediction scheme adaptive to gradients defined in four directions. The proposed prediction scheme is based on the idea that the causal pixel in the direction of least gradient value contributes maximum in prediction. Before entropy encoding, the residual errors obtained are grouped on the basis of maxplane coding which further enhances coding efficiency. The proposed work is compared with basic DPCM technique and state of the art CALIC scheme. Experimental results show compression ratio for proposed method for medical images on average is 9.65% and 30.38% better than the CALIC scheme and basic DPCM method respectively while bit rates for proposed method is 6.51% and 30.86% better than CALIC scheme and DPCM method respectively.

No abstract