A Novel Image Component Transmission Approach to Improve Image Quality and Energy Efficiency in Wireless Sensor Networks

Transcription

1 Journal of Comuter Science 3 (5: , 2007 ISSN Science Publications A Novel Image Comonent Transmission Aroach to Imrove Image Quality and nergy fficiency in Wireless Sensor Networks Wei Wang, Dongming Peng, Honggang Wang, Hamid Sharif Deartment of Comuter and lectronics ngineering, University of Nebraska-incoln, USA Abstract: Due to resource constraints, digital image transmissions are a significant challenge for image sensor based Wireless Sensor Networks (WSN. In this aer, we roose a novel robust and energy-efficient scheme, called Image Comonent Transmission (ICT in WSN by roviding various levels of reliability during image transmissions. Different comonents in the embedded wavelet image bit stream are considered and essential comonents of the comressed image such as the information for the ositions of significant wavelet coefficients are transmitted with higher quality assurance. elatively less imortant comonents such as the information for the values of ixels are transmitted with lower overhead. Simulation results show the roosed scheme achieves higher energy efficiency in WSN, with image transmission quality enhanced considerably by u to 6dB in terms of the Peak Signal to Noise atio (PSN. Keywords: Image Comonent Transmission, Wireless Sensor Network, Discrete Wavelet Transform. INTODUCTION ecent advances in image rocessing, adio Frequency (F and micro-electronic technologies have brought enormous develoment of image transmission over Wireless Sensor Networks (WSN [1]. Unlike usual image transmission designs in wireless environment, energy efficiency has the highest riority in WSN. Usually, image data is bulk sized and contains a lot of redundancy [2], which leads to challenges in designing energy efficient image transmission schemes over WSN. Wavelet based image comression techniques such as zerotree coding [3-4] and BCOT [5] achieve high comression ratio and are embedded multirate in nature, roviding unequal imortance attributes. Unequal imortance here means different arts of the comressed bit streams that exhibit different ercetual and structural imortance [2]. Imortant arts rovide critical information for reconstruction of the original image, whereas other arts may rovide fine information in higher resolution [6]. The information of natural digital image is generally conveyed by the ositions of image ixels of various values. There are different fault-tolerance levels for disturbance on the comressed image bits to be transmitted in wireless networks, articularly wireless sensor network environment. Wavelet based image comression roduces shae and osition information of the regions (e.g. how small magnitude coefficients are clustered, as well as the lighting information in regions (e.g., how large magnitude coefficients are valued. Transmission errors in shae and osition information lead to high difficulties for reconstructing the original image, while errors in the magnitudes are more tolerable during transmission. Our aroach can be exressed in two asects, i.e., (1 Identification of the essential and imortant comonents, and (2 eliable and efficient transmission of imortant arts. In the first ste, wavelet coefficients are roduced after Discrete Wavelet Transform (DWT on the original image as in the main-stream comression algorithms, and embedded bitstreams are roduced after comression. In the embedded bitstreams, multile imortant arts are identified and marked associated with their imortance levels. In the second ste, unequally imortant transmissions are alied to different comonents in the comressed image bit stream. More reliable transmission of the imortant arts enhances image quality, while less effort is ut on unimortant arts, leading to energy efficiency. Overall seaking, by exloring unequal imortance attribute of comressed image bits, our Corresonding Author: Wei Wang, Deartment of Comuter and lectronics ngineering, University of Nebraska - incoln at Omaha Camus,oom 313A, Peter Kiewit Institute1110 South 67th Street, Omaha, N 68182, USA 353

2 roosed Image Comonent Transmission (ICT aroach achieves high energy efficiency while enhancing image transmission quality in WSN. Some of the revious researches have been erformed to exlore the unequal imortance image transmission over WSN for energy efficiency. However, most of them focus on the multi resolution levels and the embedded rogressive characteristics of wavelet based image comression. How to energy-efficiently transmit image over WSN by exloring unequal imortance nature of osition information (i.e. ixel structure information and value information (i.e. ixel magnitude information is not extensively discussed in literature. The research in [2] rooses a joint source channel coding aroach for energy efficient JPG 2000 image transmission in WSN, by alying different error resilient coding rotection to different levels or layers of bitstreams. This aroach extensively exlores the multi resolution nature of bitstreams; however, the unequal imortance between structure information and magnitude information is not fully identified. In [9] multile bitstreams are created to constrain error roagation in a sub-tree, and hierarchical unequal error rotections are alied. Again, this aroach is still image-layer (or level, hierarchy oriented, without considering the unequal imortance between the structure information and magnitude information. esearches in [1] and [10] investigate the tradeoff between energy consumtion and image quality. By balancing comutational energy consumtion and transmission energy through image comression, energy efficient transmission in WSN is achieved. Those aroaches aly uniform rotection to the whole image, and do not look into the difference between the structure information and magnitude information. To the best of our knowledge, the unequal imortance nature of comressed image bits, in terms of the structure (i.e., ixel osition, shae and the magnitude (i.e., ixel value information, is not fully studied in state-of-the-art image transmission schemes over wireless networks, articularly WSN where resources (e.g., energy costs, channel qualities, etc. are severely constrained. On the other hand, such unequal imortance would rovide extraordinary oortunities for energy-efficient and robust image transmission schemes over WSN. In this aer, the roosed ICT scheme takes this novel view and synthesizes it with WSN design in a cross-layer anorama, which has the major difference from revious works. J. Comuter Sci., 3 (5: , nergy Consumtion Of Multirate Image Comonent Transmission: There have been some researches on suorting and interacting multirate signal rocessing alications in the WSN design. In our revious work on the multirate signal rocessing latform over WSN [7-8], the average energy consumtion of transmitting and receiving a data acket can be exressed in quation (1. kt = _ (1 The robabilities and energy consumtions for,,, acket transmissions, ed or successful, are exressed below. B is the desirable bit error ratio; is the acket length; is the robability of a single node sending acket, N is the neighbor count of that node; P and P denote the ower required for transmitting or receiving, resectively. T is the transmission time, and T denotes the time out value of receiving _ timeout acket. Detailed descritions can be referenced in [8]. ( ( N = 1 1 B 1 (2 = P T (3 _ timeout The robability of, and acket ure, and the corresonding energy consumtion can be exressed in quations (4 - (5: ( 1 ( B = ( 1 1 (4 = P T + P T (5 _ timeout The acket error rate due to acket ure, and the corresonding energy consumtion can be exressed in quations (6 - (7: ( 1 ( B = ( 1 (1 1 (6 = P T (7 Similarly, the acket error rate due to acket ure, and the corresonding energy consumtion can be exressed in quations (8 - (9:

3 J. Comuter Sci., 3 (5: , 2007 ( 1 ( B = ( 1 (1 (1 1 (8 = P T (9 _ timeout Above analysis gives the energy enalty and corresonding robability of acket loss events. The robability of successfully transmission of a acket, and the corresonding energy consumtion can be exressed in quations (10 - (11: = (1 (1 (1 (1 (10 = P T + P T+ P T (11 Similarly, the statistical energy consumtion of receiving a acket can be exressed in quation (12- (15: = P T (12 timeout _ = P T + P T (13 _ timeout = P T + P T (14 _ = P T+ P T+ P T+ P T (15 _ The above equations (1-(15 give the exression of acket level energy consumtion, and the arameters in quation (1 can be exressed by quations (2-(15. Furthermore, quation (2-(15 can be exressed as functions of desirable B. In quation (3, (5, (7, (9, (11, (13, (14, and (15 the transmission ower can be exressed in terms of the desirable B. According to our research ublished in [7-8], the desirable transmission ower can be determined by the B requirement, channel attenuation, modulation schemes and transmission rate. For BPSK, QPSK and QAM modulation schemes the relationshi between desirable transmission ower and desirable B is shown in quations (16-17: BPSK, QPSK 1 2 N0 Ps = s b [ erfc ( 2 B ] (16 A 2 1 M QAM N0 Ps s b ( b 1 erfc (17 = 1 B 3 2 b A From above analysis on acket level energy consumtion in quations (1-(17, we can observe that the energy consumtion of transmitting and receiving a acket can be taken as a function of desirable B. Given the factors of acket length (, overhead acket length (,,, acket transmission time (T, T, T, T, which can be solved by roviding transmission data rate and the corresonding 355 acket length, collision condition ( and N, receiving ower, and channel conditions (N0 and A, quation (1 can be rewritten in a simler form, i.e., an exlicit function of desirable B: kt = f ( B (18 The relationshi between acket level energy consumtion and B in quation (18 is monotonous because all the factors in quation (1 are monotonously increasing when B increases. Once exressed with acket level energy consumtion, the whole energy consumtion of transmitting and receiving each image can be modeled in a comonentbased scheme. For ICT, the comressed image is segmented into several comonents with unequal imortance in terms of the structure information and magnitude information. According to acket level energy consumtion shown in quation (1, we can get comonent level energy consumtion exressed in quation (19 and (20. quation (19 denotes the energy consumtion of imortant comonent containing structure information, and quation (20 denotes the energy consumtion of unimortant comonent containing magnitude information. In quation (19, Hi denotes the size of the comonent containing structure information at level i, i denotes the transmission frame length for that comonent. kt is B related arameters. Here Bi is the desirable B requirement for comonent containing structure information in level i. Biv is for comonent containing magnitude information. H i i = kt ( B i (19 i H iv iv = kt ( B iv (20 iv By summing u all the energy consumed by each comonent at each level of comression, the total energy consumtion of N-level bitstream transmission can be exressed in quation (21. N 1 H i, j = kt ( B i, j (21 i = 0 i, j where j {, v}, denoting whether this comonent containing structure information (i.e. osition or magnitude (i.e., value information. quation (21 gives the close form exression of energy consumtion for image comonent transmission. Again, the total energy consumtion for comonent based image transmission is solely determined by desirable B requirements of each comonent, as long as given the factors of comonent size, transmission frame size, and the transmission rate. In addition, if the

4 J. Comuter Sci., 3 (5: , 2007 desirable B is reduced for some comonents, then total energy consumtion is reduced due to the monotonic function of energy consumtion and desirable B requirement for each comonent. One contribution of ICT is that it rovides flexible mechanisms to design B distribution vector B=[B0, B0v, B1, B1v,,.. B(n-1, B(n-1v] (each element corresonding to the desirable B requirement assigned to the comonent], in order to achieve robustness and energy efficiency. This aroach takes advantage of the inherent inequality of image data imortance within comressed bitstreams. By scaling down B requirement of unimortant comonents, energy consumtion is significantly reduced without ercetible image quality loss. On the other hand, for transmission of an image with the same energy consumtion, using comonent aroach has better image quality, because more effort is ut on imortant comonents to assure image quality, while less effort is ut on unimortant art to achieve energy saving. Proosed ICT Aroach: ssential Comonent Identification And P-V Based Unequal rror Protection (UP: Wavelet based comression algorithms generally outut the structural information in addition to the ixel magnitude coefficients [6]. For examle, the SPIHT algorithm in [4] encodes the data sets that corresond to trees of coefficients. Although we take tree based comression algorithms as examles in this aer, the roosed image comonent transmission methodology is generic and indeendent of secific wavelet image comression algorithms. It can be easily extended to other wavelet based image comression algorithms, because essential comonents are inheritance of the original image itself. As long as the image comression algorithms can roduce unequal imortance of their outut streams consisting of ixel osition (e.g., how small wavelet coefficients are clustered and distributed and value information (e.g., how large wavelet coefficients are valued searately, they can always be alied to the roosed ICT framework. Figure 1 shows the wavelet coefficients and trees of wavelet coefficients after DWT is alied to the original image. The embedded outut bitstreams after comression is shown in Figure 2, starting with rough image (i.e., coarsest resolution followed by interleaving of tree structure information and magnitude information at various resolution levels. For examle, the H0 symbols denote information bits of tree structure in the highest wavelet coefficient level, while the H1v symbols denote the information bits of magnitudes in the second highest level. In the roosed aroach, all the tree structure information bits Hi (i=0,1, n are searated with different imortance levels after the comonent identification rocess. Unequal imortance is alied to structure information and magnitude information in different levels of wavelet coefficients, leading to the unequal imortance of transmission that will be described in next subsection. Fig. 2: mbedded outut streams of interleaving tree structure information ( and magnitude information (v. ssential comonents can be identified from comressed embedded bitstreams. In this subsection, we discuss how to energy-efficiently and robustly transmit those bitstreams in an unequal imortance rotection aroach. Figure 3 shows the data flow of such ICT framework. Fig. 3: Image Comonent Transmission Data Flow in WSN Fig. 1: eftmost: original image; Middle: wavelet coefficients; ightmost: tree structure of wavelet coefficients 356 In wavelet image comression algorithms, wavelet coefficients are comressed into embedded bitstreams after DWT is alied onto raw image. The bitstreams rocessed after essential comonents identification take the form shown in Figure 2, in which all the

5 comonents are identified, and the corresonding imortance is to be exloited during transmission. Unequal imortance transmissions are alied to those bitstreams over wireless channel, with desirable transmission ower sulied. At the sink node, bitstreams are reconstructed and decomressed into wavelet coefficients, and raw image is reconstructed by alying inverse DWT. Instead of transmitting the whole comressed image uniformly with equal imortance, different desirable B requirements are alied to different comression comonent. To exloit unequal imortance transmission, a comonent segmentation table is created in the form of Table 1 at alication layer. ach comonent in the embedded bitstreams has an entry in this segmentation table, followed by a desirable B requirement and offset osition in the bitstream. The desirable transmission ower is now shown in this segmentation since it can be calculated from desirable B requirements, with the channel attenuation factor, symbol rate and modulation scheme [7-8]. Table 1: Comonent Based Segmentation Table. The Offset byte is used for lower MAC-PHY layer to identify each comonent. Offset bytes include layer overhead from alication layer through network layer to MAC-PHY Comonent Desirable B Offset Byte H0 B0 O0 H0v B0v O0v H(n-1 B(n-1 O(n-1 H(n--1v B(n-1v O(n-1 The desirable B requirement is assigned according to the imortance of that comonent. Generally, comonents containing structure information has higher desirable B requirement than comonents containing magnitude information. For all those comonents containing structure information, they have a descending imortance sequence if the wavelet resolution level increases. This is also alicable for unimortant comonents containing magnitude information. Such concet is based on the fact that lower resolution of image data is more critical than high resolution to assure basic transmission quality requirement in WSN. The following equations show this general guideline of desirable B sequence: Bi, Bi, v,0 i n 1 (22 B i, B j,,0 i j n 1 (23 Bi, v B j, v,0 i j n 1 (24 Figure 4 shows the ICT rocess. ach comonent J. Comuter Sci., 3 (5: , is acked by MAC-PHY with the knowledge of desirable B requirement and the offset of that comonent. The comonents in segmentation table are in the same order as the embedded bitstream, starting with H0 and ending with Hnv. Therefore, MAC-PHY can identify each comonent for transmission by reading the offset byte in the segmentation table. Comonents in the outut bitstream are scanned and transmitted in sequence, at the corresonding variable B and desirable transmission ower, instead of being transmitted with uniform B and uniform ower. The transmission ower must be dynamically adjustable according to the desirable B requirement, current channel attenuation factor and modulation scheme arameters. In this aer we focus on single rate transmission, and the modulation scheme arameters are not variables of desirable transmission ower. Desirable transmission ower can be determined through quation (16 or (17, given desirable B and channel attenuation factor. Channel attenuation factor can be acquired via - handshake, and detailed descrition can be referenced in [8]. Fig. 4: Image Comonent Transmission Process in Wireless Sensor Networks with ower otimization and desirable B aroximation. Instead of treating the comressed image bit stream by a universal desirable B requirement, the ICT scheme alies different desirable B requirements to different image comonents. Given the same other factors, the energy consumtion of transmitting each image comonent is only related to comonent size, desirable B requirement and the actual transmission rate. The number of comonents and the size of each comonent are determined by the comression rocess of wavelet coefficients. SIMUATION In this section, we show the effectiveness of ICT scheme in achieving energy efficiency while enhancing image transmission quality. The digital image is transmitted between sensor nodes, by alying uniform rotection and transmission [1, 10], layer oriented

6 J. Comuter Sci., 3 (5: , 2007 transmission [2, 9], and the roosed comonent based transmission schemes. Performance data such as energy consumtion and the reconstructed image PSN values are recorded and comared. In our simulation, we emloy T-MAC [11] as MAC rotocol. The evaluation metrics are defined as the following. For T-MAC data ackets in TinyOS [12], the MAC header is 11 bytes and the ayload is 36 bytes. For the control acket such as, the length is 13 bytes. and ackets are both 15 bytes. The TA value of T-MAC used in the simulation is the same as the time of transmitting a acket. Preamble length in hysical layer is 18 bytes [13]. The receive ower is fixed to 0.01mW, and the slee ower is mW, while idle ower is mw, according to the same arameters scales in [14]. Channel attenuation factor A is -90db. The noise ower density N is the roduct of 0 23 Boltzmann constant and equivalent noise temerature Tn. Assume the noise temerature is normal room temerature 290K, and the noise ower 21 density value 4 10 J/Hz. Figure 5 shows the relationshi between the acket level energy consumtion and the desirable B requirement value. ICT scheme, each comonent in the wavelet comressed image bit stream is assigned a corresonding desirable B requirement value shown in Table 2. All the comonents in the first column of Table 2 are groued as a vector corresonding to the original image. Three scenarios including low noise (S1, medium noise (S2 and strong noise (S3 channel environments are simulated and studied. In each scenario, three transmission aroaches uniform aroach, layered aroach, comonent (i.e., ICT based aroach are analyzed. ach aroach has three desirable B vectors in those scenarios in Table 2. Table 2: Desirable B requirements vector assignment for the comonent vector in comressed image stream. ach column corresonds to a desirable B requirement vector. 3 scenarios corresonding to different channel noise levels are studied for all the three aroaches Com Uniform B Vector ayered B Vector Comonent B Vector Scen S1 S2 S3 S1 S2 S3 S1 S2 S3. H0 1e-3 5e-3 1e-2 1e-4 1e-3 1e-3 1e-4 1e-3 4e-3 H1 1e-3 5e-3 1e-2 1e-4 1e-3 1e-3 1e-4 1e-3 4e-3 H2 1e-3 5e-3 1e-2 1e-4 1e-3 1e-3 1e-4 2e-3 4e-3 H3 1e-3 5e-3 1e-2 1e-3 2e-3 1e-2 1e-4 2e-3 4e-3 H4 1e-3 5e-3 1e-2 1e-3 4e-3 1e-2 1e-4 3e-3 4e-3 H5 1e-3 5e-3 1e-2 1e-3 5e-3 1e-2 1e-4 3e-3 5e-3 H6 1e-3 5e-3 1e-2 1e-3 5e-3 1e-2 1e-3 3e-3 6e-3 H7 1e-3 5e-3 1e-2 1e-3 5e-3 1e-2 1e-3 5e-3 7e-3 H8 1e-3 5e-3 1e-2 1e-3 5e-3 1e-2 1e-3 5e-3 8e-3 H9 1e-3 5e-3 1e-2 1e-3 5e-3 1e-2 1e-3 5e-3 8e-3 H10 1e-3 5e-3 1e-2 1e-3 5e-3 1e-2 1e-3 5e-3 8e-3 H11 1e-3 5e-3 1e-2 2e-3 5e-3 2e-2 1e-3 5e-3 8e-3 H0v 1e-3 5e-3 1e-2 1e-4 1e-3 1e-3 1e-3 6e-3 8e-2 H1v 1e-3 5e-3 1e-2 1e-4 1e-3 1e-3 1e-3 6e-3 8e-2 H2v 1e-3 5e-3 1e-2 1e-4 1e-3 1e-3 1e-3 7e-3 1e-1 H3v 1e-3 5e-3 1e-2 1e-3 2e-3 1e-2 1e-3 8e-3 1e-1 H4v 1e-3 5e-3 1e-2 1e-3 4e-3 1e-2 1e-3 8e-3 1e-1 H5v 1e-3 5e-3 1e-2 1e-3 5e-3 1e-2 1e-3 9e-3 1e-1 H6v 1e-3 5e-3 1e-2 1e-3 5e-3 1e-2 1e-2 9e-3 1e-1 H7v 1e-3 5e-3 1e-2 1e-3 5e-3 1e-2 1e-2 9e-3 1e-1 H8v 1e-3 5e-3 1e-2 1e-3 5e-3 1e-2 1e-2 1e-2 1e-1 H9v 1e-3 5e-3 1e-2 1e-3 5e-3 1e-2 1e-2 1e-2 1e-1 H10 v 1e-3 5e-3 1e-2 1e-3 5e-3 1e-2 1e-2 1e-2 1e-1. Fig. 5: Average nergy Consumtion er Packet with desirable B requirements. Fig. 6: Original image with 8bits er ixel and 128*128 ixels transmitted between sensor nodes. Figure 6 shows an original image to be comressed and transmitted between sensor nodes. As roosed in 358 Figure 7, 8 and 9 show the image transmission quality comarisons among different aroaches for low noise scenario S1, medium noise S2 and strong noise S3, resectively. In each, the three aroaches are comared with erformances in terms of the image transmission quality and energy consumtion results. The reconstructed images with the roosed comonent based transmission aroach (i.e., ICT aroach in index (c in those three figures achieve the best image transmission quality, while maintain the least energy consumtion. With ICT aroach, more effort is ut on those more imortant comonents (i.e., data, which leads to higher image quality; while less effort is ut on those unimortant comonents (i.e., v data, which reduces energy consumtion. The increasing of image transmission quality by utting more effort on essential comonents incurs transmission energy overhead; however, as long as the decreased energy consumtion

7 due to less effort ut on unimortant comonents overcome that overhead, image quality imrovement and energy efficiency are both achieved. J. Comuter Sci., 3 (5: , 2007 (a (b (c Fig. 9: Comarison of reconstructed images for strong noise scenario S3. Black blocks in these figures are the results of transmission errors on imortant comonents ( data, which lead to errors with regard to the ixel ositions. Translucent blocks are due to the errors with regard to the ixel values. (a Uniform aroach, PSN = dB, energy consumtion = mJ; (b ayered aroach, PSN = dB, =0.5711mJ; (c Comonent aroach, PSN = dB. =0.5602mJ. Figure 10 shows the image quality at different channel conditions, for all the three aroaches. With all the other factors being the same, the roosed ICT aroach has higher image transmission quality than uniform aroach and layered aroach. Figure 11 shows the erformance relationshi between energy consumtion and image quality. By exloring the unequal imortance nature of various comonents in the comressed image data, the roosed ICT aroach is effective to reduce energy consumtion while enhancing image quality. Fig. 7: (a (b (c Comarison of reconstructed images for low noise scenario S1: (a Uniform aroach, PSN = dB, energy consumtion = mJ; (b ayered aroach, PSN = dB, =0.8601mJ; (c Comonent aroach, PSN = dB. =0.8288mJ. Fig. 10: econstructed image quality at different channel conditions. Fig. 8: (a (b (c Comarison of reconstructed images for medium noise scenario S2. Pictures are blurred because of the increased channel noise. Comonent aroach works better than the other two since the imortant comonents are robustly rotected. (a econstructed Images with uniform aroach, PSN = dB, nergy consumtion =0.7564mJ; (b ayered aroach, PSN=21.494dB, = mJ; (c Comonent aroach, PSN = dB, = 0.742mJ. Fig. 11: nergy consumtions at different image quality level 359

8 J. Comuter Sci., 3 (5: , 2007 CONCUSION In image sensor based WSN, visual information are obtained via a large number of medium-quality lowcost image sensor nodes. In such a system, digital images are transmitted in noisy channels from multile source sensors under resource constraints, and the visual field information is collected through large number of medium- or low-quality images catured by the source sensors. One of the critical challenges to such system design is the tradeoff of image transmission quality with energy efficiency. In this aer, we roosed a novel Image Comonent Transmission aroach to enhance image transmission quality in WSN. Our contribution lies in two asects (1 the identification of comonents with various imortance levels in comression bit streams and (2 a ractical roosal for different transmission strategies adative to different comonents. The comressed image bit stream is comonentized and differently transmitted. The unequal imortance of comonents for image reconstruction is exlored. The imortant comonents which contain bit structure information with regard to the osition of image ixels are more reliably rotected and transmitted, while unimortant art is relatively less rotected. Simulations results show that the roosed ICT aroach increases the image transmission quality considerably on noisy channels, while at the same time it achieves the energy efficiency. FNCS 1..Ferrigno, S.Marano, V.Paciello, A.Pietrosanto, Balancing comutational and transmission ower consumtion in wireless image sensor networks. I International Conference on Virtual nvironments, Human-Comuter Interfaces and Measurement Systems, 1 (1: W.Yu, Z.Sahinoglu, A.Vetro, nergy efficient JPG 2000 image transmission over wireless sensor networks. I Global Telecommunications Conference., 5 (1: J.M.Shairo, mbedded image coding using zerotrees of wavelet coefficients. I Transactions on Signal Processing., 41 (12: A.Said, W.A.Pearlman, A new, fast, and efficient image codec based on set artitioning in hierarchical trees. I Transactions on Circuits and Systems for Video Technology., 6 (3: D.Taubman, High erformance scalable image comression with BCOT. I Transactions on Image Processing., 9 (7: H.Cheng, X.i, Partial encrytion of comressed images and videos. I Transactions on Signal Processing., 48 (8: W. Wang, D. Peng, H.Wang, H. Sharif, Study of an nergy fficient Multi ate Scheme for Wireless Sensor Network MAC. In: The 2nd ACM International Worksho on QoS and Security for Wireless Networks, W.Wang, D.Peng, H.Wang, H.Sharif, T.Wysocki, B.Wysocki, An nergy fficient MAC-PHY Aroach to Suort Distributed Source Coding in Wireless Sensor Network. In: The 5th Worksho on the Internet, Telecommunications and Signal Processing. 9. M.Wu, C.Chen, Multile bitstream image transmission over wireless sensor networks. In: Proceedings of I Sensors, C.Chiasserini,.Magli, nergy consumtion and image quality in wireless video-surveillance networks. In: The 13th I International Symosium on Personal, Indoor and Mobile adio Communications T. van Dam, K. angendoen, An adative energy efficient MAC rotocol for wireless sensor networks. In: Proceedings of the First ACM Conference on mbedded Networked Sensor Systems. P htt:// 13. I. hee, A. Warrier, M. Aia, J. Min, Z-MAC: a hybrid MAC for wireless sensor networks. In: Proceedings of the 3rd international conference on mbedded networked sensor systems, J. Polastre, J. Hill, and D. Culler, Versatile low ower media access for wireless sensor networks. In: Proceedings of the Second ACM Conference on mbedded Networked Sensor Systems