Browsing by Author "Wu, BF"

Now showing 1 - 20 of 58

A discrete wavelet transform codec design
(2004-12-01)
Hu, YQ
;
Wu, BF
;
Su, CY
;
電控工程研究所
;
Institute of Electrical and Control Engineering
This manuscript presents a VLSI architecture and its design rule, called embedded instruction code (EIC), to realize discrete wavelet transform (DWT) codec in a single chip. Since the essential computation of DWT is convolution, we build a set of multiplication instruction, MUL, and the addition instruction, ADD, to complete the work. We segment the computation paths of DWT according to the multiplication and addition, and apply the instruction codes to execute the operators. Besides, we offer a parallel arithmetic logic unit (PALU) organization that is composed of two multipliers and four adders (2M4A) in our design. Thus, the instruction codes programmed by EIC control the PALU to compute efficiently. Additionally, we establish a few necessary registers in PALU, and the number of registers depends on the wavelet filters' length and the decomposition level. Yet, the numbers of multipliers and adders do not increase as we execute the DWT or the inverse DWT (IDWT) in multilevel decomposition. Furthermore, we deduce the similarity between DWT and IDWT, so the functions can be integrated in the same architecture. Besides, we schedule the instructions; thus, the execution of the multilevel processes can be achieved without superfluous PALU in a single chip. Moreover, we solve the boundary problem of DWT by using the symmetric extension. Therefore, the perfect reconstruction (PR) condition for DWT requirement can be accomplished. Through EIC, we can systematically generate a flexible instruction codes while we adopt different filters. Our chip supports up to six levels of decomposition, and versatile image specifications, e.g., VGA, MPEG-1, MPEG-2, and 1024 x 1024 image sizes. The processing speed is 7.78 Mpixel/s when the operation frequency, for normal case, is 100 MHz.
A discriminant analysis based recursive automatic thresholding approach for image segmentation
(2005-07-01)
Wu, BF
;
Chen, YL
;
Chiu, CC
;
電控工程研究所
;
Institute of Electrical and Control Engineering
In this study, we have proposed an efficient automatic multilevel thresholding method for image segmentation. An effective criterion for measuring the separability of the homogenous objects in the image, based on discriminant analysis, has been introduced to automatically determine the number of thresholding levels to be performed. Then, by applying this discriminant criterion, the object regions with homogeneous illuminations in the image can be recursively and automatically thresholded into separate segmented images. The proposed method is fast and effective in analyzing and thresholding the histogram of the image. In order to conduct an equitable comparative performance evaluation of the proposed method with other thresholding methods, a combinatorial scheme is also introduced to properly reduce the computational complexity of performing multilevel thresholding. The experimental results demonstrated that the proposed method is feasible and computationally efficient in automatic multilevel thresholding for image segmentation.
A fast convolution algorithm for biorthogonal wavelet image compression
(1999-03-01)
Wu, BF
;
Su, CY
;
電控工程研究所
;
Institute of Electrical and Control Engineering
Symmetric filters and symmetric extension of image edges have been widely used in wavelet image compression. Since the filters are symmetric, it is possible to take advantage of the symmetric property to reduce the computational complexity for the filtering. In this paper, we present a fast convolution algorithm for the discrete wavelet transform (DWT) and the inverse DWT (IDWT) such that the transform time can be greatly reduced. Compared with regular convolution, the new algorithm can decrease the multiplication operations by nearly one half. Converted into real programming, it sped up the DWT and IDWT in our experiments by at least 12% and 55%, respectively. Incorporated with enhancing zerotree coding, the proposed algorithm results in a rapid and efficient coder. Experimental results showed that the coder is competitive with other high performance coders. The pro posed convolution algorithm is also suitable for many types of wavelet-based coding, including wavelet video coding.
A fast lane and vehicle detection approach for autonomous vehicles
(2005)
Wu, BF
;
Lin, CT
;
Chen, CJ
;
Lai, TC
;
Liao, HL
This paper proposes a lane detection algorithm, a fuzzy-based vehicle detection approach and a distance measurement method for intelligent autonomous vehicles. The lane detection algorithm reduces computational load and has proven successful on highway and in urban settings at different velocities (110km/h and 50km/h, respectively). The proposed fuzzy-based vehicle detection approach named Contour Size Similarity (CSS), compares the contour size of objects by fuzzy rules. CSS efficiently distinguishes target obstacles, namely vehicles, from other objects detected and differentiates between obstacles and pattern on the road surface. After the nearest vehicle in the lane is found, we estimate its distance to the intelligent autonomous vehicle according to its position in the image. Then the information obtained would sever as aids for automatic driving of the intelligent autonomous vehicles.
A fuzzy vehicle detection based on contour size similarity
(2005)
Wu, BF
;
Lin, CT
;
電控工程研究所
;
Institute of Electrical and Control Engineering
In this paper, the obstacle detection for autonomous vehicle is carried out by a monocular camera, which is capable of detecting multiple obstacles, identifying the difference between obstacles and text or shadows on the road surface The task of detection can be fulfilled under the circumstances of general brightness or in the presence of strong sunshine. An obstacle detection approach named Contour Size.Similarity (CSS) is developed in this paper. CSS is a way to make use of, the contour size of objects, and compare the similarity between objects and obstacles detected by means of fuzzy so as to judge whether the objects in the images are the obstacles we want to detect. The I.obstacles in this paper mainly refer to vehicles. Vehicles and obstacles with similar contour of vehicles can be detected by CSS. Then, we use section division to find out the nearest obstacle in separate section to provide necessary information for autonomous vehicles.
A high-performance and memory-efficient pipeline architecture for the 5/3 and 9/7 discrete wavelet transform of JPEG2000 codec
(2005-12-01)
Wu, BF
;
Lin, CF
;
電控工程研究所
;
Institute of Electrical and Control Engineering
In this paper, we propose a high-performance and memory-efficient pipeline architecture which performs the one-level two-dimensional (2-D) discrete wavelet transform (DWT) in the 513 and 9/7 filters. In general, the internal memory size of 2-D architecture highly depends on the pipeline registers of one-dimensional (I-D) DWT. Based on the lifting-based DWT algorithm, the primitive data path is modified and an efficient pipeline architecture is derived to shorten the data path. Accordingly, under the same arithmetic resources, the 1-D DWT pipeline architecture can operate at a higher processing speed (up to 200 MHz in 0.25-mu m technology) than other pipelined architectures with direct implementation. The proposed 2-D DWT architecture is composed of two 1-D processors (column and row processors). Based on the modified algorithm, the row processor can partially execute each row-wise transform with only two column-processed data. Thus, the pipeline registers of 1-D architecture do not fully turn into the internal memory of 2-D DWT. For an N x M image, only 3.5N internal memory is required for the 513 filter, and 5.5N is required for the 9/7 filter to perform the one-level 2-D DWT decomposition with the critical path of one multiplier delay (i.e., N and M indicate the height and width of an image). The pipeline data path is regular and practicable. Finally, the proposed architecture implements the 513 and 9/7 filters by cascading the three key components.
A low memory QCB-based DWT for JPEG2000 coprocessor supporting large tile size
(2005)
Wu, BF
;
Lin, CF
;
電控工程研究所
;
Institute of Electrical and Control Engineering
JPEG 2000, which provides a higher compression ratio than the traditional JPEG, is an upcoming compression standard for still images. The experimental results imply that larger tile size used for JPEG 2000 results in better image quality. However, processing the large tile image requires more memory in the hardware implementation. To reduce the hardware resources, a QCB (quad code-block)-based DWT method is proposed to support the process of large tile image with low memory. Based on the QCB-based DWT engine, three code-blocks belonging to LH(0), HL(0) and HH(0) bands can be generated after each fixed time slice recursively and the EBC (embedded block coding) processors can directly process the three code-blocks. It can save the size of tile memory up to 75%. Moreover, the remaining 1/4 size of tile memory can be decreased through the zero holding extension for the unavailable data. That is, it only requires 24K bytes memory to support the process of 512x512 tile image, with slight image degradation, especially at low bit rates. The low memory requirement makes the hardware implementation practicable.
A low memory zerotree coding for arbitrarily shaped objects
(2003-03-01)
Su, CY
;
Wu, BF
;
電控工程研究所
;
Institute of Electrical and Control Engineering
The Set Partitioning In Hierarchical Trees (SPIHT) algorithm is a computationally simple and efficient zerotree coding technique for image compression. However, high working memory requirement is its main drawback for hardware realization. In this study, we present a low memory zerotree coder (LMZC), which requires much less working memory than WHIT. The LMZC coding algorithm abandons the use of lists, defines a different tree structure, and merges the sorting pass and the refinement pass together. The main techniques of LMZC are the recursive programming and a top-bit scheme (TBS). In TBS, the top bits of transformed coefficients are used to store the coding status of coefficients instead of the lists used in SPIRT. In order to achieve high coding efficiency, shape-adaptive discrete wavelet transforms are used to transformation arbitrarily shaped objects. A compact emplacement of the transformed coefficients is also proposed to further reduce working memory. The LMZC carefully treats "don't care" nodes in the wavelet tree and does not use bits to code such nodes. Comparison of LMZC with SPHIT shows that for coding a 768 x 512 color image, LMZC saves at least 5.3 MBytes(1) of memory but only increases a little execution time and reduces minor peak signal-to noise ratio (PSNR) values, thereby making it highly promising for some memory limited applications.
A low-complexity MP3 algorithm that uses a new rate control and a fast dequantization
(2005-05-01)
Yen, CH
;
Lin, YS
;
Wu, BF
;
電控工程研究所
;
Institute of Electrical and Control Engineering
This work presents a low-complexity MP3 algorithm over a fixed-point arithmetic. A new rate control is introduced for the MP3 encoding algorithm, rather than the rate control in ordinary MP3. The computational complexity of the rate control is reduced by taking the loop-independent components outside the loop and accelerating the nonuniform quantizer using a hybrid scheme. The hybrid scheme includes a lookup-table method for smaller numbers and a linear piecewise approximation for larger numbers. A precise method for predicting the quantizer parameter is developed to decrease the number of times the rate control is execute. hybrid scheme is also used in MP3 decoding algorithm to accelerate the dequantization. However, the approximation for larger numbers is two-tier. The first tier is a linear piecewise approximation that yields a rough value. The second tier uses the rough value as the initial value of the first -order Newton's method to obtain a more closely approximated value. The precise method for prediction has a statistically hit rate of 43%, and the new rate control consumes no more than 4.5 MIPS. The proposed dequantization consumes no more than 2.38 MIPS, and has an error-to-signal ratio of under 0.012%. The implementation Of the complexity-reduced MP3 algorithm over 16bit fixed-point arithmetic is subjectively tested to evaluate the quality of the complexity-reduced MP3 algorithm.
A real-time multiple-vehicle detection and tracking system with prior occlusion detection and resolution
(2005)
Wu, BF
;
Lin, SP
;
Chen, YH
;
電控工程研究所
;
Institute of Electrical and Control Engineering
The proposed multiple-vehicle detection and tracking (MVDT) system utilizes a color background to segment moving objects and exploits relations among the moving objects and existed trajectories to track vehicles. Initially, the background is extracted by classification. Then, it is regularly updated by previous moving objects to guarantee robust segmentation in luminance-change circumstance. For partial wrong converged background due to roadside parking vehicles. it will be corrected later by checking fed back trajectories to avoid false detection after the vehicles moving away. In tracking processing, the relations of distances or distances and angles are applied to determine whether to create, extend, and delete a trajectory. If occlusion detected after trajectory creation, it will be resolved by rule-based tracking reasoning. Otherwise, lane information will be used. Finally, traffic parameter calculations based on the trajectories art, listed. Moreover, for easy setup, parameter automation for the system is proposed.
A real-time robust lane detection approach for autonomous vehicle environment
(2004)
Wu, BF
;
Chen, CJ
;
Chiu, CC
;
Lai, TC
;
電控工程研究所
This paper proposes a real-time lane detection algorithm which provides adaptive region of interesting to reduce computational load and is robust for real road environment, e.g., road inclination, road width calibration. Compared to previous approaches, the flat road assumption is released in our approach so that the result is more close to real road environment. This work has been successfully verified on highway and urban with different velocities 110 km/h and 50 km/h, respectively, in the real vehicle, TAIWAN iTS-1, which is an experimental car of Taiwan ITS project. The algorithm is also demonstrated on the lanes which are with shadows, texts, crooks, or sheltered by other vehicles for sunny and night environment.(1)
A rescheduling and fast pipeline VLSI architecture for lifting-based discrete wavelet transform
(2003)
Wu, BF
;
Lin, CF
;
電控工程研究所
;
Institute of Electrical and Control Engineering
In this paper, we propose a fast pipeline VLSI architecture for ID lifting-based discrete wavelet transform (DWT). This design method merges the filtering steps called the predictor and updater into one single step. Based on this modified algorithm, we explore the data dependency of the input and output signals, and thus make the pipeline design more efficiently for hardware implementation under the same processor elements proposed in previous works. Moreover, the inverse DWT case also adopts the same architecture as the forward DWT. Finally, the area and the working frequency of the proposed architecture in 0.35um technology are 2.511 x 2.510 mm(2), and 150 MHz, respectively.
Adaptive self-quantization in wavelet-based fractal image compression
(1999-05-01)
Wu, BF
;
Hu, YQ
;
Hsu, HH
;
電控工程研究所
;
Institute of Electrical and Control Engineering
Finding a model to quantize the scale factors in wavelet-based fractal image compression is a complicated issue. To avoid error, it is helpful to model the distribution of the scale factors and quantize them before the computation process by iterated function systems. Traditionally, a fixed model with uniform distribution was frequently adopted. This is not sophisticated enough, however, to quantize these scale factors from errors since, in general, the factors are not uniformly distributed. We propose an adaptive algorithm with self-quantization to overcome this drawback. Except for the functions of adaptation and self-quantification, the approach has the optimal property that the fundamental objective is to reduce the quantization errors.
Algorithms for compressing compound document images with large text/background overlap
(2004-12-01)
Wu, BF
;
Chiu, CC
;
Chen, YL
;
電控工程研究所
;
Institute of Electrical and Control Engineering
Two algorithms are presented for compressing image documents, with a high compression ratio for both colour and monochromatic compound document images. The proposed algorithms apply a new method of segmentation to separate the text from the image in a compound document in which the text overlaps the background. The segmentation method classifies document images into three planes: the text plane, the background (non-text) plane and the text's colour plane, each of which are processed using different compression techniques. The text plane is compressed using the pattern matching technique, called JB2. Wavelet transform and zerotree coding are used to compress the background plane and the text's colour plane. Assigning bits for different planes yields high-quality compound document images with both a high compression ratio and well presented text. The proposed algorithms greatly outperform two well known image compression methods, JPEG and DjVu, and enable the effective extraction of the text from a complex background, achieving a high compression ratio for compound document images.
An efficient architecture for JPEG2000 coprocessor
(2004-11-01)
Wu, BF
;
Lin, CF
;
電控工程研究所
;
Institute of Electrical and Control Engineering
JPEG2000 is a new international standard for still image compression. It provides various functions in one single coding stream and the better compression quality than the traditional JPEG, especially in the high compression ratio. However, the heavy computation and large internal memory requirement still restrict the consumer electronics applications. In this paper, we propose a QCB (quad code block)-based DWT method to achieve the higher parallelism than the traditional DWT approach of JPEG2000 coding process. Based on the QCB-based DWT engine, three code blocks con be completely generated after every fixed time slice recursively. Thus, the DWT and EBCOT processors can process simultaneously and the high computational EBCOT then has the higher parallelism of the JPEG2000 encoding system. By changing the output timing of the DWT process and parallelizing with EBCOT, the internal tile memory size can be reduced by a factor of 4. The memory access cycles between the internal tile memory and the code block memory also decrease with the smooth encoding flow(1).
An efficient hierarchical motion estimation on algorithm and its VLSI architecture
(2005)
Wu, BF
;
Peng, HY
;
Chen, CJ
;
Yu, TL
;
電控工程研究所
This manuscript addresses the design and implementation of an efficient hierarchical motion estimation algorithm, HMEA, using multiple motion vector candidates to increase the estimation quality. HMEA satisfies the high estimation performance and efficient VLSI implementation. The algorithm adopts multi-resolution frames to reduce the computation complexity, and the original frame is down sampled into one-fourth and one-sixteenth respectively. At the smallest resolution, the least two motion vector candidates are selected using full-search block matching algorithm with search range equals to 4. At the middle level, these two motion vector candidates are used as center points for local searches, and the search range is reduce to 2. Then, at the original resolution level, the final motion vector is obtained from local search around the single candidate from the middle level. This paper also describes an efficient VLSI architecture based on the HMEA. We implement this architecture with about 59K gates and 798 bytes of RAM for a search range of [-16.0, +15.5] by using a synthesizable VHDL.
An efficient VLSI implementation of the discrete wavelet transform using embedded instruction codes for symmetric filters
(2003-09-01)
Wu, BF
;
Hu, YQ
;
電控工程研究所
;
Institute of Electrical and Control Engineering
This work presents a VLSI design rule, namely, an embedded instruction code (EIC), for the discrete wavelet transform (DWT). Our approach derives from the essential computations of DWT, and we establish a set of multiplication instructions, MUL,, and the addition instruction, ADD. In addition, we propose a parallel arithmetic logic unit (PALU) with two multipliers and four adders, called 2M4A. With these requirements, the DWT computation paths can be calculated more efficiently with limited PALUs. Furthermore, since the EIC is operated under the PALU, the number of needed inner registers depends on the wavelet filters' length. Besides, the boundary problem of DWT has also been resolved by the symmetric extension. Moreover, the two-dimensional inverse DWT (2-D IDWT) can be completed using the same PALU for 2-D DWT; the only changes needed to be made are the instruction codes and coefficients. Our chip supports up to six levels of decomposition and versatile image specifications, e.g., VGA, MPEG-1, MPEG-2 and 1024 x 1024 image sizes.
An encrypted mobile embedded surveillance system
(2005)
Wu, BF
;
Peng, HY
;
Chen, CJ
;
Chan, YH
;
電控工程研究所
This manuscript addresses the design and implementation of an encrypted mobile embedded surveillance system, EMESS, including a wireless server and mobile devices. Not only the mobility of the hand-held devices, but also the surveillance system can work in a moving body, then, a real mobile surveillance system is provided. In the security applications of vehicles, EMESS also can provide the images of thieves and robbers to the vehicle owner. We complete the related works of EMESS and the corresponding application programs for wireless communication of hand-held devices.
An error-correcting stream cipher design with state-hopping architecture
(2005-01-01)
Yen, CH
;
Wu, BF
;
電控工程研究所
;
Institute of Electrical and Control Engineering
A new architecture of stream cipher based oil state-hopping shift registers and a pseudorandom vector generator (PRVG) is introduced. The proposed Stream cipher merges secrecy coding and channel coding into one processing Step. It could be either a pure cryptosystem or a secrecy-channel coding by demand. Considered as cryptography, the PRVG generates pseudo random vectors which are treated as keystreams setting Lip the encryption scheme. This is different from the general concept in stream ciphers, state-hopping shift registers do not generate a pseudo random sequence but act as substitutions on plaintexts. Viewed as channel coding. the state-hopping shift registers play as the ones in convolution code and the PRVG generates a sequence of pseudorandom vector to determine the Trellis diagram. If the system acts as a pure cryptosystem. the decoding scheme is exactly the inverse of the encryption Scheme. When the error-correcting ability is chosen. a modified sequential decoding is proposed.
An innovation approach to optimize a Kalman filter with an H-infinity error bound by secant method
(1996-07-01)
Wu, BF
;
Hsu, HH
;
交大名義發表
;
電控工程研究所
;
National Chiao Tung University
Designing a Kalman filter with a constraint on the H-infinity norm of the estimation error was first developed by Bernstein and Haddad in 1989. The main result is a sufficient condition for characterizing the Kalman filter. In this paper, similar to the standard Kalman filter, the properties of orthogonal principles are also shown to be preserved. Furthermore, the uniqueness, as opposed to an H-infinity filter, of the filter is implied by the orthogonal principles. An innovative approach to obtaining the minimum energy with a constraint on the H-infinity norm of the estimation en ol is proposed since the original work of Bernstein and Haddad does not, in general, reach the minimum energy of the estimation error. By means of the Secant method, the energy of the estimation error can be reduced as much as possible, under the condition that the H-infinity error bound is still satisfied.