Shachi Natu

Transform based face recognition with partial and full feature vector using DCT and Walsh transform

This paper presents, efficient transform based face recognition technique which considers full and partial feature vector of an image. 2D-DCT and Walsh transform is applied on the resized image of size 128x128, to obtain its feature vector. Partial feature vector is obtained by selecting 75% rows and columns of feature vector, 50% rows and columns of feature vector and so on. The smallest size of partial feature vector is selected as 4x4. Proposed technique is tested on two different databases. Georgia Tech Face Database contains JPEG color images and Indian Face Database contains Bitmap color images of varying size. Recognition rate is calculated for varying size of selected feature vector using DCT and Walsh transform and compared. Also computational complexity in terms of number of CPU units is compared in both the cases: with full feature vector and with partial feature vector. Results show that, Walsh transform gives better recognition rate than DCT and number of CPU units required using 2D- Walsh transform is almost 9 times less than that of required by using 2D-DCT. This is because the multiplications required in Walsh transform are zero.

Performance Comparison of Speaker Identification Using DCT, Walsh, Haar on Full and Row Mean of Spectrogram

ABSTRACT This paper aims to provide different approaches to text dependent speaker identification using various transformation techniques such as DCT, Walsh and Haar transform along with use of spectrograms. Set of spectrograms obtained from speech samples is used as image database for the study undertaken. This image database is then subjected to various transforms. Using Euclidean distance as measure of similarity, most appropriate speaker match is obtained which is declared to be identified speaker. Each transform is applied to spectrograms in two different ways: on full image and on Row Mean of an image. In both the ways, effect of different number of coefficients of transformed image is observed. Further, comparison of all three transformation techniques on spectrograms in both the ways shows that numbers of mathematical computations required for Walsh transform is much lesser than number of mathematical computations required in case of DCT on spectrograms. Whereas, use of Haar transform on spectrograms drastically reduces the number of mathematical computation with almost equal identification rate. Transformation techniques on Row Mean give better identification rate than transformation technique on full image.

Biometric watermarking using partial DCT-Walsh wavelet and SVD

Biometric has ever increasing demand in the field of security. However, integrity and security of biometric used to authenticate a legal user are the inherited problems of it. Since watermarking hides one form of digital data into other, its use in securing biometric data gives an added advantage. This concept has led to biometric watermarking technique in which one biometric can be embedded into another biometric feature of user to enhance security. In this paper a robust biometric watermarking technique using face images as host and iris images as watermark has been proposed with the combination of hybrid wavelet and Singular Value Decomposition. DCT-Walsh hybrid wavelet transform is used in its partial form (column transform) on host and watermark images. Low frequency coefficients of partially transformed hosts are subjected to SVD. Singular values of partially transformed watermark are adaptively scaled and are embedded in these singular values of host. Robustness of proposed technique is evaluated against compression, selective and random cropping, noise addition and resizing attack. Average absolute pixel difference between embedded and extracted watermark is used as similarity measure between them. Proposed technique is observed to be highly robust against these attacks except selective cropping. However this limitation is eliminated when iris images are embedded in middle frequency coefficients of host.

Halftone Image Data Compression Using Kekre’s Fast Code Book Generation (KFCG) Algorithm for Vector Quantization

Halftone technique is well known for printing where binary data is required. 8:1 compression ratio is achieved by half toning method. To get higher compression ratio the same technique can be used in image processing. In our earlier work different half toning operators are proposed. The half toning operator of size 3X3 which effectively take only one tap operation. Hence the computational complexity and memory space is reduced. For further compression of the image data Vector Quantization technique is used. Vector Quantization technique itself gives very high compression ratio. To avoid time and computational complexity Kekre’s Fast Code Book Generation (KFCG) algorithm is used. In this paper we have used 8, 16, 32, 64, 128 and 256 codebook sizes are used. The pixel group of 2X2 size is used. The experimental results are obtained for various images. For image data compression and image quality measurement, we have used Compression Ratio and Mean Square Error as measuring parameters respectively. We have got result with good Compression Ratio and acceptable image quality. This proposed combination of two compression technique is suitable for video data streaming, where low bit rate for data transmission is the major constraint.

Robust watermarking by SVD of watermark embedded in DKT-DCT and DCT wavelet column transform of host image

Watermarking in wavelet domain and with SVD is popular due to its robustness. In this paper a watermarking technique using DCT wavelet and hybrid DKT-DCT wavelet along with SVD is proposed. Wavelet transform is applied on host and SVD is applied on watermark. Few singular values of watermark are embedded in mid frequency band of host. Scaling of singular values is adaptively done for each channel (Red, green and blue) using the highest transform coefficient from selected mid frequency band and first singular value of corresponding channel of watermark. Singular values of watermark are placed at the index positions of closely matching transform coefficients. This along with the adaptive selection of scaling factor adds to the robustness of watermarking technique. Performance of the proposed technique is evaluated against image processing attacks like cropping, compression using orthogonal transforms, noise addition, histogram equalization and resizing. Performance for DCT wavelet and DKT-DCT wavelet is compared and in many of the attacks DCT wavelet is found to be better than DKT-DCT wavelet.

Performance Evaluation of Digital Color Image Watermarking Using Column Walsh Wavelet Transform

This paper proposes a robust watermarking technique using wavelet transform generated from well-known orthogonal transform Walsh. Watermark embedding is done in middle frequency band of column wavelet transformed host image. Performance of proposed technique is evaluated against image processing attacks like compression, cropping, addition of run length noise with binary and Gaussian distribution and image resizing. Comparison of performance of these transforms is done on the basis of robustness to attacks using Mean Absolute Error (MAE) as a metric of robustness. Column wavelet is found preferable over full wavelet. Also column Walsh wavelet is preferable over column DCT wavelet for robustness.

Robust Watermarking by Embedding Watermark in Sorted Mid-frequency Coefficients of Column Transform using DKT-DCT Wavelet

In this paper we introduce a watermarking method which is based on hybrid wavelet transform generated from two orthogonal transforms namely Discrete Kekre Transform and Discrete Cosine transform. As introduced in our previous work, column transform of host and watermark is obtained to make the technique computationally efficient. Watermark is compressed and normalized before embedding to improve imperceptibility of watermarked image. Middle frequency host transform coefficients are used to embed the watermark. Instead of random (row wise) embedment of watermark coefficients into host coefficients, sorting is applied to both coefficients to have a maximum match or minimum difference between them while embedding. Sorting improves the performance of the technique when compared to previous work done without sorting.