Abstract-Recently, a two-phase scheme for removing saltand-pepper impulse noise has been proposed [14]. In the first phase, an adaptive median filter is used to identify pixels which are likely to be contaminated by noise (noise candidates). In the second phase, the image is restored by minimizing a specialized regularization functional that applies only to those selected noise candidates. As an extension of this work, we propose an efficient method to accomplish the second phase. The speed of our method can be double as that of the method proposed in [14] for images contaminated by 30% salt-and-pepper noise and will be faster for higher noise level.

Mobile devices use embedded processors with low computing capabilities to reduce power consumption. Since floatingpoint anthmetic units are power hungry, computationàlly intensive jobs must be accomplished with either digital signal processors or hardware co-processors. In this paper, we propose to perform fixed-point arithmetic on an integer hardware unit. We illustrate the advantages ofour approach by implementing fmgerprint verification on mobile devices Keywords: embedded system, fixed-point arithmetic, fingerprint verification, smart card, personal digital assistant . IntroductionTo enhance the security of a mobile device (cellular phone, smart card, or personal digital assistant), active participation in the authentication procedure is a must. Emerging authentication methods (fingerprint recognition, handwritten signature, and graphical password) emphasize non-integer computation, in contrast to the traditional personal identification number (PIN) based solely on integer operations. However, most ofthe currently available mobile devices possess only integer hardware units. Moreover, the trend of multi-application usage of mobile devices necessitates a more sophisticated computational mechanism. These demands motivate our study of domain specific support for fixedpoint computation. Mobile Devices and Their CharacteristicsSmart cards come in various types. Briefly, a smart card [6] is shaped and sized like a credit card, and is embedded with special hardwired logic to securely retain (and perhaps process) critical information. It offers good protection on the client side, when online activities such as Internet e-commerce are conducted. Although there are smart cards with only memory, we focus on smart cards with processors. To be specific, we will consider Java cards equipped with 5 12 bytes of RAM as yolatile storage, 32K bytes of EEPROM as non-volatile memory, and 32-bit RISC processors [1]. On the software side, we want the processors to conform to the Java Card 2.0 specification [2] . On the hardware side, some expensive features may be absent. For example, the hardware can perform integer but not floating-point arithmetic. Herein lies the problem of the feasibility of the in-card processor to conduct the entire fingerprint matching process, which involves many decimal calculations.An embedded processor is a dedicated real-time machine providing control and computation as a part of a complete system [3]. It contains a microprocessor chip (or chips) that controls the operation of displays, motors, lights and a range of other devices [4]. Examples of embedded systems are personal digital assistants and mobile phones. Unlike generalpurpose processors, embedded processors are designed and optimized for specific classes of applications [51. Due to many limitations (small on-chip memory and low computation power), it is difficult to port PC-based software to embedded systems. Thus, implementing the computationally intensive fingerprint verification algorithm by software on such systems will pose a big challen...