1 August 2007 Design of a fast vector quantization image encoder
Author Affiliations +
Optical Engineering, 46(8), 087008 (2007). doi:10.1117/1.2768066
Vector quantization (VQ) is an efficient technique for signal compression. However, it requires much encoding time to find the closest codeword for every input vector. We propose a fast encoding method to speed up the encoding. With the help of a table that is created off-line and can be used by all the images, the encoder searches only part of the entire codebook. The proposed method is implemented to encode Lena and other images to test its performance. Compared to full-searching VQ (FS-VQ), although the encoder searches only about 20 codewords in the codebook for every input vector, more than 95% of the codewords searched by the proposed method are the same as the results searched by FS-VQ on average. In addition, we also adopt partial distortion searching (PDS) and lookup table (LUT) to decrease the mathematic computation. This saves 98.44% of the encoding time and 98.07% of the mathematic operation while encoding Lena. The proposed method is superior to all the existing fast VQ encoding methods. While encoding 100 nature images for testing, it can save more than 97% of the encoding time and mathematic operations, but the PSNR decays at most only 0.19 dB, which is invisible to human eyes.
Yung-Gi Wu, "Design of a fast vector quantization image encoder," Optical Engineering 46(8), 087008 (1 August 2007). http://dx.doi.org/10.1117/1.2768066

Computer programming

Image quality




Optical engineering

Image compression


Open source database of images DEIMOS extension for large...
Proceedings of SPIE (September 23 2014)
Optical Neural Networks : System And Device Technologies
Proceedings of SPIE (February 08 1988)
Optical design of the ESPRESSO spectrograph at VLT
Proceedings of SPIE (July 14 2010)
Comparison Exam: Impact On PACS
Proceedings of SPIE (September 16 1985)
Advanced Helmet Mounted Display (AHMD)
Proceedings of SPIE (May 04 2007)

Back to Top