Search

With the development of the ability to crack encrypted images, conventional image encryption techniques are no longer safe enough. Disguising a to-be-encrypted image into another visually-different image that is similar to the prepared target image is a solution that can be used to transmit an image securely. Inspired by the existing reversible image transformation technique, in this paper, we propose an improved method for camouflaging images that further decrease the distortion between the final camouflaged image and the target image. First, the G channel of the secret image is transformed and further refined to create a tentative camouflaged G channel with the reference of the G channel of the target image. Then, the color difference channels of the secret image are transformed with the reference of subtraction of the R channel (or B channel) and the tentatively camouflaged G channel. After shifting the color difference channels back to the R channel (or B channel), the sub-blocks are refined by a 16-candidate-pattern optimization strategy to generate the tentative camouflaged R channel (or B channel). After the combination of the RGB channels to generate the tentative camouflaged image, the final camouflaged version is generated by embedding all of the auxiliary information, which is collected for lossless recovery of the secret image, into the tentative image in a reversible manner. The experimental results demonstrated the efficacy of the proposed method, and our average gain in the color image peak signal-to-noise ratio (CPSNR) was more than 0.35 dB, whereas the state-of-the-art method is around 32.28 dB.

In this paper, we proposed a bit-rate adjustable color image compression technique based on block truncation coding. To exploit the similarity among the neighboring pixels, the quadtree segmentation technique is used to divide the color image into variable-sized blocks based on their block activities. Different rules are used to encode the image blocks of different sizes. Experimental results reveal that good image qualities of the reconstructed color images can be achieved by the proposed technique while keeping low bit rates.

To enhance the signal integrity of high-density neural-sensing signals, this work presents an implantable high spatial resolution μ-probe array with through-silicon via (TSV) 2.5D integration technology that realizes a miniaturized implantable device on flexible printed circuit (FPC) interposer. The proposed microsystem was composed of two 32-channel neural sensing chips and one radio frequency chip for neural signal processing. The μ-probe array can achieve better signal-to-noise ratio with neural-signal acquisition and processing circuit composed of a pseudo-resistor-based analog front-end amplifier. Moreover, a receiving antenna is also implemented on the backside of FPC for wireless data and power transmission. The feasibility of the proposed μ-probe array, Tx and Rx antenna, 32-channel neural sensing circuits in the 64-channel wireless microsystem have been successfully demonstrated for future integration and animal experiments.

For implanted neural-sensing devices, one of the remaining challenges is to transmit stable power/data (P/D) transmission for high spatiotemporal resolution neural data. This paper presents a miniaturized implantable 128-channel wireless neural-sensing microsystem using TSV-embedded dissolvable μ-needle array, a flexible interposer and 4 dies by 2.5D/3D TSV heterogeneous SiP technology. The 4 dies are 2 neural-signal acquisition ICs implemented by 90nm CMOS, 1 neural-signal processor by 40nm CMOS and 1 wireless P/D transmission circuitry by 0.18μm CMOS. Thus, the proposed wireless microsystem realizes 128-channel neural-signal sensing within the area of 5mm × 5mm, neural feature extraction and wireless P/D transmission using an on-interposer inductor. The overall average power of the circuits in this microsystem is only 9.85mW.

To cater complex topography sensors and circuit chips for smart lifestyle applications using Internet of Things (IoT), we develop a novel efficient double-self-assembly packaging approach by 3D/2.5D heterogeneous integration. This technology is able to integrate various kinds of processed chips, especially with uneven surface and bottom topography. With liquid surface tension, complex topography chips can be handled and self-aligned on precise locations. Double-self-assembly approach demonstrates temporary bonding on carrier wafer and self-aligned permanent bonding in a pressure sensing system packaging. The optimized volume of liquid to use on various complex topography shapes and areas are investigated, while Cu/In low temperature metal bonding and self-assembly approach for high I/O density and low bonding resistance are demonstrated. Excellent electrical characteristics and mechanical strength of show the feasibility of the double-self-assembly approach on complex topography surface chips packaging applications.

A breakthrough in vertical stacking and bonding method in flexible substrates is developed and presented in this paper. Unlike current ACF and NCP approaches, which are used in most manufacturing industry, our novel stacking method uses only metal thin film as bonding material. Two different bonding material bumps are carried out in our experiment. Both show better electrical and reliability outcomes compared to current methods. Our proposed method can reach ultra-fine pitch, metal bonding without polymer material, simplified manufacturing process, better electrical performance and less reliability issues. It can be considered as a future bendable and flexible substrate stacking approach.

Local brain connectivity is expected to lead to new models for neurological diseases, which may in turn result in advanced understanding and better treatment. This paper presents a neural sensing microsystem integrated with TSV-embedded dissolvable µ-needles array, ENIG bonding technology, biocompatible Au-TSV flexible interposer and neural recording circuits, for neural sensing implantation. An ultra-thin film bonding approach is proposed for integration of interposer assembly. Removing bonding wire by proposed bonding technology, the dimension of neural sensing system can be minimized to reduce surgical area and promote implant success rate. The signal quality of neural recording can be significantly improved by eliminating complex signal paths with 2.5D TSV integration from neural sensing interface to neural recording circuits.

An innovative bonding approach applied to flexible substrate bonding is proposed in this research. Thin-film metal pad is fabricated as bonding medium to provide vertical stacking. As compared to other bonding methods, such as anisotropic conductive film (ACF) or non-conductive paste (NCP), thin-film metal pad bonding approach provides the most attractive advantage of fine-pitch. The bonding condition at low temperature exhibits good electrical performance, proving its great potential in the applications of flexible substrate stacking.

Highly integrated neural sensing microsystems are crucial to capture accurate signals for brain function investigations. In this paper, a 256-channel/25 mm2 neural sensing microsystem is presented based on through-silicon-via (TSV) 2.5D integration. This microsystem composes of dissolvable μ-needles, TSV-embedded μ-probes, 256-channel neural amplifiers, 11-bit area-power-efficient SAR ADCs and serializers. Based on the dissolvable μ-needles and TSV 2.5D integration, this microsystem can detect 256 ECoG/LFP signals within the small area of 5mm × 5mm. Additionally, the neural amplifier realizes 57.8dB gain with only 9.8μW for each channel, and the 9.7-bit ENOB of the SAR ADC at 32kS/s can be achieved with 0.42μW and 0.036 mm2. The overall power of this microsystem is only 3.79mW for 256-channel neural sensing.

In the past decades, image processing technologies have been applied to process medical images. Usually, image segmentation is an important strategy. Fuzzy c-means clustering algorithm has been wildly used for segmentation of brain magnetic resonance image. In the paper, we implement a genetic Fuzzy c-means clustering algorithm based on embedded graphic process units system, NVIDIA TK1, to accelerate computation speed of time-consuming on segmenting brain magnetic resonance image. The experimental results show that the proposed algorithm not only can used to analyze such image on cheap device but also gains from the performance.

In this paper, a simple process for high yield CMOS-compatible and heterogeneous integrated chip-to-chip structure without TSV is demonstrated. This scheme provides two chips consisted of the shortest interconnect path by Cu/Sn pillar bump and electroless nickel immersion gold (ENIG) pad bonding. One of the key technologies of 3D integration process is bump plating on the uneven topography. Since passivation layer covers the periphery of the top metal layer, subsequent electroplating process resulted to the increase in height of bump edge which is higher than bump center resulting in concave shape. A new and unique plating solution was developed to solve the issue during the electroplating pillar bump. Basic electrical characteristics including resistance and current leakage were investigated with reliability tests. The stable reliability tests results and excellent electrical performance show that the 3D heterogeneous integration structure is potentially applicable for 3D applications in the future.

In this paper, a 3D chip-to-chip hetero-integration scheme without Si interposer is demonstrated under 270°C thermal budget. The scheme successfully integrates advanced and common technology node logic chips by Cu/Sn to ENIG bumping. Cu/Sn μ-bumps are electroplated on common technology node and ENIG joints are electroless-plated on advanced technology node opening pads, respectively. Herein, 60 μm bump pitch, 40 μm diameter of Cu/Sn μ-bump and 50 μm diameter of ENIG are presented. Without cracks and voids, the 3D C2C scheme gives an efficient approach for future development of 3D IC.

In this paper, a novel crystal component package based on silicon TSV interposer substrate is demonstrated by using 3D integration technologies. It is distinct from conventional crystal component using ceramic-based substrate. This crystal component with advanced silicon-based substrate shows great manufacturability to replace traditional fabrication approach. In addition, the SU-8 sealing bonding structure provides the possibility to substitute for the conventional metal lid with reliability investigation. Finally, a novel silicon-based crystal component package with 3D integration is successfully fabricated through general semiconductor techniques to replace traditional manufacture.

Nowadays, with the explosive growth of the network technologies many new applications and services have been developed on Internet. World Wide Web can provide these services provided without the limitation of time and location. Obviously, the number of user is dramatically increasing from amount of the visitations of web pages. In our previous work, we proposed an algorithm to discover more significant information from visited web pages to provide this information to web designers or policy makers to adjust the presentation of their Web contents. However, this algorithm is time-consuming approach due to it needs to scan the whole database many times. Therefore, we propose a GPGPU-based Preference Utility algorithm to enhance the performance by GPGPU parallel model. The proposed algorithm is developed on NVIDIA CUDA architecture. The experimental results show that the proposed method can achieve about 7x times over CPU-based method. The proposed algorithm can used to mine the information from web log data efficiently.

In this paper, one novel packaging approach for crystal resonator with the quartz crystal is demonstrated, developed and characterized. The proposed crystal resonator of the novel package adopts several 3-D core technologies, such as Cu TSVs, thin-film Cu/Sn eutectic bonding, and wafer thinning. A 1210 crystal resonator with quartz blank mount is successfully developed with Cu TSVs that enables the electrical connection of the signal electrode on the quartz blank. The quartz blank is protected by the simultaneous formation of strength reinforcement and excellent sealing capacity of thin-film Cu/Sn eutectic bonding. With the well fabrication and the enhanced structure, excellent leakage results in the MIL-STD-883 hermetic test (8.5*10-9 ~1*10-8 Pa m3/sec by He leak detector) show the crystal resonator using 3D integration has the great performance and manufacturability to replace the conventional metal or ceramic enclosures for the next generation products.

In this paper, a detailed examination on TSV and cavity inductive coupled plasma (ICP) etching is presented. We investigated the relation such as etching loop number, TSV etching depth and etching rate. Due to particles knocked off from the hard mask and then fallen down to the TSV and cavity bottom, micro-masking issue becomes serious after ICP etching. In addition, parameters of isotropic etching, pressure, and RF bias were studied to investigate the process of micro-masking removal.

Bosch reactive ion etching is widely used for TSV formation. The micro-masking formed during etching can be successfully removed by adjusting the internal parameters during etching. The smooth high-aspect-ratio TSVs were further developed in wafer-level fabrication. Finally, a two-step etching process was developed to achieve tapered TSVs for the following Cu plating process.

Baculovirus has emerged to be a promising gene delivery vector and adipose-derived stem cells have captured interests for bone regeneration. Massive segmental defects arising from trauma or tumor resection remain a challenging clinical problem. To heal massive, segmental bone defects using adipose derived stem cells (ASCs), which alone cannot heal large defects, we hypothesized that sustained expression of factors promoting bone regeneration (BMP2) and angiogenesis (VEGF) provides continuous stimuli to augment the healing. Insect baculovirus (BV) holds promise for gene therapy and efficiently transduces stem cells, but it only mediates transient transgene expression. Therefore we developed a dual BV system whereby one BV expressed FLP recombinase (BacFLP) while the other hybrid BV harbored an Frt-flanking transgene cassette. After sequential transduction with BacFLP and the hybrid BV, FLP/Frt-mediated recombination occurred in up to 46% of ASCs, leading to cassette excision off the BV genome, formation and persistence of episomal transgene and prolongation of expression to >28 days. Transduction of ASCs with the BMP2-encoding hybrid BV led to prolonged BMP2 expression and augmented ASCs osteogenesis even without other osteogenic supplements. ASCs engineered by the hybrid vectors mediating sustained BMP2/VEGF expression healed the critical-size (10 mm) segmental bone defects in 12 out of 12 New Zealand White rabbits in 8 weeks, which significantly outperformed ASCs engineered with BV transiently expressing BMP2/VEGF with respect to healing rate, bone metabolism, bone volume, bone density, angiogenesis and mechanical properties. These data confirmed our hypothesis that persistent BMP2/VEGF expression is essential. The use of the hybrid BV vector represents a novel therapy to treat massive segmental defects necessitating concerted ossification and vascularization.

Establishing criteria for knowledge management (KM) is important, because criteria help to establish a basis for assessing the value and evaluating its results. More importantly, the criteria will tell us what the KM outcomes are and their relevance to organizational performance. The literature review has revealed that widely accepted criteria and performance measures have not been developed for KM. Delphi Technique and survey-based research using a questionnaire targeting KM professionals as respondents were aimed at establishing criteria for assessing KM success for different types of organizations. The results show what organizations consider important outcomes of a KM initiative. Contributions from this research effort should support government, nonprofit, and for-profit organizations in making decisions about KM initiatives and measuring KM efforts in terms of its relevance to the performance of organizations. Future research efforts can focus on developing these KM outcomes into detailed measures.

In this paper, we propose a new color image retrieval scheme based on color features. The goal of this scheme is to effectively retrieve images similar to a query image in a color image database. In this scheme, a color image is first divided into three planes. Then the local features of the given image can be found by calculating the color difference of each plane. Thereafter, the color difference co-occurrence matrix (CDCOM) of each color image can be derived through these local features. The CDCOM is regarded as the global feature of the color image. Consequently, CDCOM is used to find images similar to the query image. Experiment results show that the proposed scheme can effectively retrieve the desired and similar images.

In this paper, the code transformation algorithms between the fixed binary linear quadtree scheme (FBLQ)and the constant bit-length linear quadtree scheme(CBLQ) are proposed. There exists no code transformation algorithm between FBLQ and CBLQ. To solve this problem that when one receives a FBLQ or CBLQ code and does not have the corresponding decoding program, with our new code transformation algorithms at hand, she/he can transform the received code into CBLQ or FBLQ code that she/he can handle.The code transformation algorithms can also be used to reduce the load an image lays on the system by transforming it from a higher complexity code to a lower one.

In this paper, we present an image-hiding scheme based on genetic algorithm. The secret messages are embedded into a compressed image of AMBTC. Genetic algorithm is developed to find the best substitution of AMBTC bitmap. The proposed scheme provides high visual quality of the stego-image. The enhanced system of the proposed scheme increases embedding capacity while retaining good quality of the stego-image. Experimental results show that the proposed scheme outperform the comparative schemes.

In this paper, a new adaptive image compression scheme based on JPEG was proposed to improve the image quality of the JPEG baseline sequential image compression codec. In this scheme, the Hough transformation technique is employed as an edge-oriented classifier to categorize image blocks into four classes. Different types of image blocks are processed by different quantization and encoding ways for achieving an effective compression performance. In order to preserve edge information, the adaptive quantization tables based on the human visual system and zig-zag scan sequences are used. For both objective and subjective visual image quality, simulation results show that the performance of the proposed scheme is comparable to that of other compared schemes.

In this paper, a watermarking-based image authentication scheme used to verify the contents of transmission grayscale images is proposed. In this scheme, the watermark of one given image is generated from both frequency domain and spatial domain. First, the number of encoding stages of each DWT coefficient during the multi-stage encoding is taken as frequency watermark. Second, the mean values of blocks are stored as spatial watermark. The watermark is then embedded into SPIHT encoded list of significant pixels (LSP) bit stream. By comparing the embedded watermark and the corresponding message extracted from decoded image to be authenticated, the given image can then be authenticated. The experimental results show that the proposed scheme can distinguish acceptable operations from malicious manipulations. Furthermore, the positions of intended modified areas can be located accurately and recovered approximately

A novel image-hiding scheme based on the vector quantization scheme is proposed in this paper. Multiple codebooks are used in this scheme and the mutual relationship between a pair of VQ indices is exploited to hide secret data. The secret data are embedded in the codebook indices corresponding to the pair of VQ indices. The hidden secret data can be extracted by decoding the VQ compressed codes. Experimental results show that the embedding capacity and image quality are superior to Jo and Kim's (2002) scheme. Besides, in the proposed scheme, the image quality based on embedding the same amount of secret data is always better than the compared scheme.

In this paper, algorithms for transforming code between bintree and quadtree are proposed. The quadtree and bintree are two commonly used representations for image manipulations. However, there exists no algorithm for transforming code between bintree and quadtree. When one receives a DF-expression of a bintree or quadtree code and does not have the corresponding decoding program, she/he, with our new code transformation algorithms at hand, can transform the received code into a quadtree or bintree code that she/he can handle. After code transformation, the receiver can see whether the output code is smaller than the input code. From it, she/he can further decide which code is more economic in its memory space needed; therefore, she/he can store the code with smaller size for saving more memory space.

A novel method based on hyperplane tests is introduced for the design of binary tree structured vector quantisation (TSVQ) codebooks. A hyperplane test is a simple technique to decide on which side of the hyperplane any given point is. This technique is used in our method to find the closest codeword for each training vector. From the experimental results, it is shown the method we propose not only designs better TSVQ codebooks but also requires less computational cost than that of the traditional scheme.

A novel two-stage image compression scheme is proposed. In the first stage, differential pulse code modulation (DPCM) is used to decorrelate the raw image data. Next, an effective scheme based on the Huffman coding scheme is proposed to encode the residual image. According to the experimental results, the newly proposed scheme could significantly reduce the cost for the Huffman table while achieving high compression performance. In other words, the newly proposed scheme provides a good means for lossless compression of high-resolution images.

In vector quantization schemes, a good performance codebook is required in both the encoding and the decoding procedures. In this paper, we propose a novel codebook design scheme using a tree structure. The study on the distribution of codewords in a codebook motivated us to design our algorithm. Generally, a good distribution of codewords will lead to good codebook performance, and vice versa. As the experimental results show, the method we propose can increase the image quality (PSNR) by about 1.214 dB and 1.240 dB with the codebook sizes 128 and 256, compared with the codebook generated by the LBG algorithm, respectively. Therefore, it is obvious that the proposed scheme provides a good approach to design better VQ codebooks for image compression.