Search

The Embedded Zerotree Wavelet (EZW) coder which can be considered as a degree-0 zerotree coder, and Set Partitioning In Hierarchical Trees (SPIHT) coder which can be considered as a degree-2 zerotree coder are two well-known image compression techniques. However, the performance of a degree-1 zerotree coder has not been well investigated. In this paper, an Embedded Descendent-Only Zerotree Wavelet (EDOZW) coder which can be considered as a degree-1 zerotree coder is proposed. The EDOZW coder is a modified version of SPIHT coder with Arithmetic Coding (SPIHT-AC). When coding the entry in the List of Insignificant Set (LIS) in SPIHT, it is necessary to first determine whether the entry is Type A or Type B. Then, a different scheme is used to handle the entry. But in contrast to SPIHT, it is no longer necessary to differentiate the type of entry in EDOZW. Only one type of entry is used in EDOZW and a recursive algorithm can be used to encode all the entries easily. This will help to reduce the overall complexity. Moreover, the EDOZW also achieved better performance at high bit rates. The EDOZW can serve as an alternative solution for image compression other than EZW and SPIHT. [ABSTRACT FROM AUTHOR]

Usage of 3D images for face recognition has become more popular in the past few years since they are not affected by pose changes. Hence, there exists a need to store or transmit 3D images efficiently. This is important when increasing amount of 3D images is needed to be stored in a limited space or sent over a fixed bandwidth channel. A possible solution is to compress the 3D image. Unfortunately, this may cause the recognition rate to drop. In this paper, a 3D face range recognition system with compression is proposed and the effect of using compressed 3D range images on the recognition rate is investigated. The Set Partitioning in Hierarchical Trees (SPIHT) coding technique is proposed to be used for the compression. This technique is an improvement of the Embedded Zerotree Wavelet (EZW) coding technique. From the simulation results, when comparing uncompressed probe images and probe images compressed using SPIHT coding, the compressed image recognition rate ranges from being lower to being slightly higher than uncompressed probe image recognition rate, depending on bit rate. This proves that a 3D face range recognition system using compressed images is a feasible alternative to a system without using compressed images and should be investigated since the benefits like smaller file storage size, faster image transmission time and better recognition rates are important. [ABSTRACT FROM AUTHOR]

A high performance 2D one-bit-transform (1BT) motion estimation algorithm with smoothing and preprocessing (S+P) is introduced in this paper. In 1BT motion estimation algorithm, the 8-bit current frame (c frame) and reference frame (p frame) are transformed into their 1-bit representation image, before calculating the Sum of Absolute Difference (SAD) and performing the search operations. For our proposed algorithm, a smoothing threshold (ThresholdS) is incorporated into the filtering kernel, which is used to perform the transformation from 8-bit image into its 1BT image. This smoothing technique can greatly reduce the scattering noise created in the 1BT image. After transformation, the 1BT image for the c frame and p frame is divided into macroblocks. The macroblock in the c frame will be first compared to the macroblock at the same position in the p frame. If the SAD is below the preprocessing threshold (ThresholdP), the macroblock is considered to have negligible movement and search operation is not required. This preprocessing technique can greatly reduce the total number of search operations. Simulation results show that an improvement up to 0.64 dB, with reduction in search operation up to 93.36% is achieved for video conferencing sequences. [ABSTRACT FROM AUTHOR]

The 3D Set Partitioning In Hierarchical Trees (SPIHT) is a video coding algorithm extended from the SPIHT algorithm, which is initially introduced by A. Said and W. Pearlman for image coding. Previous works have show that the performance of 3D SPIHT with Arithmetic Coding (AC) is comparable to H.263 and MPEG-2. Moreover, the 3D SPIHT algorithm also supports progressive video transmission with controllable bit rates. This paper presents a new configuration of adaptive arithmetic model (also known as adaptive model) which can enhance the coding efficiency of AC, and achieves better performance in terms of Peak Signal-to-Noise Ratio (PSNR). The adaptive model is a very important module in AC, which is used to store the probability distribution of all the symbols that appear in a system. In the new configuration, each type of output bits in 3D SPIHT is assigned with a separate set of adaptive models. This new configuration takes into account the different probability patterns which exist in each type of output bits. On the other hand, the maximum frequency used to reset the adaptive model is also presented. This is a parameter that determines the adaptation rate of the AC. The simulation results show that the new configuration can improve the mean PSNR by 0.2 to 1.66 dB for various video test sequences in QCIF and SIF formats. [ABSTRACT FROM AUTHOR]