Takahiko Masuzaki

JPEG2000 adaptive rate control for embedded systems

A novel rate control scheme is proposed dedicatedly for JPEG2000 image coding. By predicting bitrate of coded data and updating it adaptively, the proposed scheme can be executed in parallel with the code-block coding of code-block coding such as coefficient bit modeling and arithmetic coding. The proposed scheme successfully reduces computational cost and working memory size of the process down to 29% and 13%, respectively, comparing to a conventional approach in case of 1/16 compression, and hence is suitable to be used in embedded systems.

Adaptive rate control for JPEG2000 image coding in embedded systems

To cope with the recent mobile scenes where images are used aggressively, a novel rate control scheme is proposed. The proposed scheme, dedicated for JPEG2000 image coding, aims at achieving low computational cost and small working memory size while maintaining high image quality. By predicting the adequate number of coding passes and updating it adaptively in code-block coding, the proposed scheme reduces both computational cost and working memory size for bitstream buffering down to 29% and 13%, respectively.

Scalable Design Framework for JPEG2000 System Architecture

For the exploration of system architecture dedicated to JPEG2000 coding, decoding and codec, a novel design framework is constructed. In order to utilize the scalability of JPEG2000 algorithm aggressively in system implementation, three types of modules are prepared for JPEG2000 coding/decoding/codec procedures, i.e. software, software accelerated with user-defined instructions, and dedicated hardware. Specifically, dedicated hardware modules for forward and inverse discrete wavelet transformation (shortly DWT), entropy coder, entropy decoder, and entropy codec as well as software acceleration of DWT procedure are devised to be used in the framework. Furthermore, a JPEG2000 encoder LSI, which consists of a configurable processor Xtensa, the DWT module, and the entropy coder, is fabricated to exemplify the system implementation designed through the use of proposed framework.

Design framework for JPEG2000 encoding system architecture

A novel design framework for exploring JPEG2000 encoding system architecture is proposed. In this framework, each procedure of JPEG2000 encoding is selectively implemented among those by software, software accelerated with user-defined instructions, or dedicated hardware, while maintaining the compliance with the requirements and constraints of each terminals and applications, so as to optimize the encoding system organization. Dedicated hardware modules for DWT and entropy coder as well as software acceleration of DWT procedure are devised to be used in the framework. Furthermore, an LSI, which consists of the DWT module and the entropy coder, is fabricated to exemplify the system implementation designed through the use of proposed framework.

High speed JPEG2000 encoder by configurable processor

This paper discusses a design of high speed JPEG2000 encoder. JPEG2000 entropy coding is realized by hardware module since its computational cost accounts for roughly 65% of total according to software profiling. Discrete wavelet transformation (DWT) is accelerated by attaching user-defined instructions to Tensilicas configurable processor Xtensa. Utilizing the 8,700 gate entropy coder with 27 Kbit of memory and the custom instructions implemented by 8,000 gates, the number of cycles needed to encode an image is reduced to 31%.

JPEG2000 high-speed progressive decoding scheme

An efficient scheme for JPEG2000 progressive decoding is proposed, which is capable of handling image codestreams with SNR progressiveness. In order to avoid processing the same codestream data more than once when decoding SNR progressive images, a pair of techniques are introduced in our decoding scheme; reusing of intermediary decoded data and differential IDWT. Comprehensive evaluation of our scheme demonstrating that with 20% increase of required memory size, more than 40% of computational costs can be reduced in comparison with conventional (non-progressive) decoding scheme.

Efficient memory architecture for JPEG2000 entropy codec

An encoding/decoding process of JPEG2000 requires much more computation power than that of conventional JPEG mainly due to the complexity of entropy encoding/decoding. Thus usually multiple entropy codec hardware modules are implemented in parallel to process entropy encoding/decoding. This module, however, requests many small-size memories to store intermediate data, and when multiple modules are implemented on a chip, employment of the large number of SRAMs increases difficulty of whole chip layout. In this paper, an efficient memory architecture of the entropy encoding/decoding module is proposed, in which three approaches are attempted by utilizing one-bank SRAMs and internal registers. As a result, the efficient memory organization for a target process technology can be explored

Design Framework For JPEG2000 System Architecture

Abstract For the exploration of system architecture dedicated to JPEG2000 coding, decoding and codec, a novel design framework is constructed. In order to utilize the scalability of JPEG2000 algorithm aggressively in system implementation, three types of modules are prepared for JPEG2000 coding/decoding/codec processes, i.e. softwaze, software accelerated with user-defined instructions, and dedicated hardware. Specifically, dedicated hazdware modules for forward and inverse discrete wavelet transformation (shortly DWT), entropy coder, entropy decoder, and entropy codec as well as software acceleration for the DWT process are devised to be used in the framework. Furthermore, a JPEG2000 encoder LSI, which consists of a configurable processor Xtensa, the DWT module, and the entropy coder, is fabricated to exemplify the system implementation designed through the use of proposed framework.

High quality Motion JPEG2000 coding scheme based on the human visual system

A novel high quality Motion JPEG2000 coding scheme is proposed, which is based on the human visual system being insensitive to the high frequency components of an image. This method enhances subjective video quality by controlling the amount of data allocated to each code-block according to the extent of its motion. The proposed method enables video coding without degrading the subjective quality of the existing method at lower bitrate.

Video quality enhancement for Motion JPEG2000 encoding based on the human visual system

A novel Motion JPEG2000 encoding method is proposed, which is based on the human visual system being insensitive to high frequency component of image. This method enhances subjective video quality by controlling bitrate of each region, such as a tile or a code-block, or each wavelet coefficient according to its amount of motion. The proposed method can encode video streams whose subjective quality is the same as that by existing method at lower bitrate.