Yu-Ting Pai

A Simple and Accurate Color Face Detection Algorithm in Complex Background

Human face detection plays an important role in many applications such as video surveillance, face recognition, and face image database management. This paper describes a fast face detection algorithm with accurate results. We use lighting compensation to improve the performance of color-based scheme, and reduce the computation complexity of feature-based scheme. Our method is effective on facial variations such as dark/bright vision, close eyes, open mouth, a half-profile face, and pseudo faces. It is worth stressing that our algorithm can also discriminate cartoon and human face correctly. The experimental results show that our approach can detect a frame in 111 msecs with the 92.3% detection rate

Adaptive thresholding algorithm: Efficient computation technique based on intelligent block detection for degraded document images

Document image binarization involves converting gray level images into binary images, which is a feature that has significantly impacted many portable devices in recent years, including PDAs and mobile camera phones. Given the limited memory space and the computational power of portable devices, reducing the computational complexity of an embedded system is of priority concern. This work presents an efficient document image binarization algorithm with low computational complexity and high performance. Integrating the advantages of global and local methods allows the proposed algorithm to divide the document image into several regions. A threshold surface is then constructed based on the diversity and the intensity of each region to derive the binary image. Experimental results demonstrate the effectiveness of the proposed method in providing a promising binarization outcome and low computational cost.

An efficient thresholding algorithm for degraded document images based on intelligent block detection

Document image binarization plays an important role in many applications such as optical character recognition, automatic bank check processing, and vehicle license recognition. This paper proposes an efficient binarization algorithm with intelligent block size detection. Based on the image characteristic, the document image is automatically divided into several blocks with various sizes. Then, a threshold surface is constructed to derive the binary image. Experimental results show that the proposed method not only provide promising binarization result, but also low computational cost.

Low Power Huffman Coding for High Performance Data Transmission

In this paper, we propose a low power Huffman coding by reducing the number of switching activity during data transmitting. The low power Huffman coding consists of the following two steps: firstly, reduce the switching activity between each symbol by changing the locations of leaves node in Huffman tree. Secondly, reconstruct the Huffman tree by properly changing interior nodes to decrease switching activities of each codeword. The experimental results show that the proposed Huffman coding algorithm can reduce switching activity up to 7.25% on average and 18.04% for the best case compared with traditional Huffman coding approach in executable files. Besides, the average power consumption reduction ratio is 3.81% for other files. As a result, the proposed algorithm is very useful for data transmissions over large self-capacitances wire

Sub-Trees Modification of Huffman Coding for Stuffing Bits Reduction and Efficient NRZI Data Transmission

In recent decades, image and video compression was widely used on network access. However, there are few researches focused on the behavior between data transmission and multimedia compression. Therefore, this paper considers this problem between the encoding of compression and transmission to develop a low bit rate transmission scheme based on Huffman encoding. The proposed method can balance “0” and “1” bits to save the issue by analyzing the probability of the miss match in the typical Huffman tree. Moreover, the proposed method also can modify the transitional tree under the same compression ratio. Experimental results show that the proposed method can reduce the stuffing bits to 51.13% of standard JPEG compression. Besides, the file size after the proposed encoding is the same with the original one. It is observed that the proposed method provides a way to reduce the transmitted bits under the same compression ratio.

A hardware-efficient color segmentation algorithm for face detection

This paper develops a hardware-efficient color segmentation algorithm that is especially suitable to implement on hardware for face detection. Since the modulized design is adopted in the proposed algorithm without floating-point operation, the computational cost is directly reduced for hardware design. The proposed algorithm consists of a color space modeling module and a feature enhancement module. The significant skin/lip color features distribution can be accurately detected by using our proposed algorithm to facilitate the face detection. The proposed algorithm was implemented on a field-programmable gate array (FPGA) system for verifying its efficiency. Compared with other state-of-the-art algorithms, the proposed algorithm can significantly decrease the computational cost of the hardware implementation by using color segmentation instead of the overall analysis of the color distribution. Experimental results have verified that our proposed FPGA system occupies only 3,202 logic cells, or about five times less than the current comparable FPGA system with better detection rate.

Energy-Efficient watermark algorithm based on pairing mechanism

In recent years, digital watermarking is a technique for labeling digital images by hiding secret information which can protect the copyright. The goal of this paper is to develop a DCT-based watermarking algorithm for low power and high performance. Our energy-efficient of technique focus on algorithm level processing. We improve Hsu and Wus algorithm by using DCT coefficients to pairing blocks directly. The experimental results show our approach not only can reduce a halt of pairing operations required, but also increase the PSNR around 0.2 db.

Honeycomb Model Based Skin Color Detector for Face Detection

Skin color is an important feature for face detection and recognition in color images. In order to obtain the possible face regions in color images, the skin color models are always constructed by statistical analysis. Owing to low accuracy of the static models, researches have discussed several dynamic models to correct input image such as illumination compensation, white balance, edge points addition, etc. Unfortunately, it is possible that some objects whose color is the same as the definition exist, and the previous methods can not separate real skin item from skin color background. Thus, to enhance skin color separation, this paper presents a honeycomb model to recognize the real human skin and the skin color items. First, the possible skin color is estimated from the pixels of database, and the honeycomb structure is built in HSV color space according to the training samples. Then, the personal skin is captured in one of the honeycomb cells. The performance of the new skin color detector technique has been tested under complex lighting source and background environments. It is observed that the proposed model can effectively improve the segmentation results. Especially, the honeycomb model is capable of separating the human face which connected with other face or skin color background.

Exploring the tradeoff between power and detection rate for a face detection algorithm

Human face detection plays an important role in many applications such as human computer interface, video surveillance, face recognition, and face image database management. Recently, most researches mainly focus on developing a high detection rate algorithm to improve the performance. Since a high performance face detection algorithm requires complicated framework, it results high energy consumption and hard to implement in portable devices whose battery lifes are always restricted. Therefore, how to develop a face detection algorithm with both promising detection rate and low power dissipation has become a crucial research topic. In this paper, we present a comprehensive analysis of energy requirements in a wide range of the most used algorithms in the face detection program. The experimental results show the power dissipation and detection rate of seven different cases. Base on the results, we conclude and discuss various opportunities for realizing energy-efficient implementations of a face detection program.

Design and Analysis of Low Power Dynamic Bus Based on RLC simulation

In this paper, we propose a low power dynamic bus encoding scheme which simultaneously reduces the capacitive and inductive effects by the measurement of real RLC model. It should be noted that our method does not need a sufficient knowledge of the patterns on the bus. Our experimental results show that the proposed approach can save power consumption of the bus up to 12% compared to the nonencoded case. We also propose an area-aware scheme to optimize our circuits in terms of power consumption and area. The scheme can reduce the circuit area up to 29% while keeping almost the same power reduction