Vector Quantization technology is a well-known compression technique. It reduces the amount of image information by replacing the original image with a template called a codebook. Because conventional approaches to extracting codebooks are based on Kohonen, LBG, and similar learning algorithms, there are problems in obtaining high compression ratios and high image quality for a broad range of images. Ohmi Laboratories has created a basic codebook that, in theory, can resolve these problems, enabling it to be adapted to a diverse range of images. Based on these fundamental techniques, I&F Co., Ltd. introduced adaptive resolution alteration processing in cooperation with Ohmi Laboratories, and established a Vector Quantization still-image compression technique capable of compressing and reconstructing still images with high compression ratios and high image quality.
The JPEG method is the dominant compression technique currently in use for still images and is based on discrete cosine transforms (DCT), the use of which can cause a type of image degradation, known as "mosquito noise," to occur. In addition, JPEG2000 and other compression techniques are based on transform coding from the spatial domain to the frequency domain, and image degradation such as mosquito noise will occur in principle when images are highly compressed. In contrast, the Vector Quantization still-image compression technique developed here, in theory, generates no mosquito noise, and with the introduction of adaptive resolution alteration processing, can also limit the generation of the block noise peculiar to Vector Quantization. The result is higher quality decompressed images compared to JPEG and JPEG2000. Particularly in images incorporating text, there is an overwhelming performance difference of 5 to 10 dB in terms of image quality. In the future, the Vector Quantization approach has the potential to become the fundamental technology for compression methods used in transmitting images over networks.
Application to Wireless LAN Data Projectors
Using the JPEG method to compress images such as presentation data for transmission to wireless LAN data projectors is problematic. It has proven difficult to maintain high image quality and high display speeds when transmitting over wireless networks because of the previously mentioned "mosquito noise," which is a significant cause of degradation in image quality.
Sharp determined that Vector Quantization still-image compression technology is ideal for application to wireless LAN data projectors. Together with Ohmi Laboratories and I&F Co., Ltd., Sharp specifically adapted this compression technique to be used in projectors operating in a wireless network LAN (IEEE 802.11b).
The new image compression algorithm and wireless network technology developed at this time not only achieves high image quality in decompressed images but, thanks to the Vector Quantization technique, also enables significantly faster image decompression processing compared to the JPEG method. This makes it possible to achieve real-time wireless presentations, a feat which previously had been difficult to achieve. In addition, the following features are also made possible:
- Quasi Animation display achieved by compressing and transmitting only the parts of the image that have changed since the previous frame.
- Many-to-one transmission (Multiple PCs to one projector)
- One-to-many transmission (One PC to the multiple projectors)
This functionality will enable new application ideas never before imagined, such as simultaneous, interactive use of data projectors by many people.
