A Hybrid Compression Technique for Segmented Hand Veins Using Quad Tree Decomposition MOHAMED NAGY SAAD Biomedical Engineering Department Misr University for Science and Technology 6 th of October, Egypt nagybioeng@yahoo.com AHMED HISHAM KANDIL Systems and Biomedical Engineering Department Cairo University Giza, Egypt ahkandil_1@yahoo.com Abstract—Biometrics are techniques for automatically identifying and authenticating an individual based on his physiological or behavioral characteristics. Hand vein is one of the biometric modalities. Hand vein check measures the shape and size of veins in the back of the hand in a grayscale image. In this paper, hybrid compression technique is applied on ninety hand vein images. This hybrid technique is combining the advantages of lossless techniques and lossy techniques. Only the essential information is selected and compressed using lossless technique, and nonessential information is compressed using lossy technique. The observed parameters are compression ratio (CR), total compression time (TCT), mean square error (MSE), and peak signal to noise ratio (PSNR). The goal is to maximize the CR while preserving images’ information. This is acheived using object segmentation procedure and quad tree decomposition (QTD) as preprocessing steps for the compression process. Applying the hybrid technique on the dataset images results in a CR in the range of 89.56%. Keywords— Discrete Cosine Transform; Hand Veins; Huffman Coding; Image Compression; Quad Tree Decomposition. I. INTRODUCTION Hand veins modality is used to verify individual identity based on the unique patterns of veins [1], [2]. Generally, the biometric images imply specific features such as uniform background, relatively large homogenous regions and high resolution [3]. In the proposed hybrid model, regions containing essential information is compressed using lossless algorithm, while other regions are compressed by lossy algorithm. This is achieved by segmenting hand veins as a preprocessing step (segmented regions), then using quad tree decomposition (QTD) function to compress the segmented regions, and to apply lossy compression technique on the other regions. The data set consists of 90 images collected by Shahin [1]. The chosen parameters: compression ratio (CR), total compression time (TCT), mean square error (MSE), and peak signal-to-noise ratio (PSNR) are measured and the resulted compressed image is verified through Hand Vein Verification program [1]. CR is the ratio of the reduction in number of bits representing the digital image after compression to the number of bits representing the original digital image as shown in (1). It is the most significant metric of performance measure of a data compression algorithm. Lossless techniques yield modest CRs, while the lossy ones yield high CRs [4].         Where CR is the calculated compression ratio, a is the original image size and b is the compressed image size. TCT is the time delay required for compressing and decompressing the image. TCT is one of the parameters that measure the performance of the compression algorithms. Complex compression algorithm requires relatively long time leading to serious problems in interactive applications [5]. PSNR is defined as shown in (2).           Where MSE is defined as shown in (3).     ∑ ∑ [( )    ( )]        Where, X is the number. of rows and Y is the no. of columns of the image. I(i,j) is the original image pixels and I’(i,j) is the reconstructed image pixels. The grayscale image pixels vary between 0 and 255 values. The PSNR is a quantitative measure for image quality evaluation [6], [7]. II. SYSTEM DESCRIPTION The testing algorithm is done using Matlab operating on Windows XP. The dataset images are 8-bit per pixel grayscale images of type (BMP), as shown in Fig.1(a). Their dimensions are 320 * 240 pixels and so image size is 76 KB. The hand vein image consists of a hand grip containing veins and image background. Essential information in this image is the veins tree (identification network). Segmenting the veins is accomplished through five steps: background removal, 162