The hardware architecture design for color transforms and the image coding with JND

• Main Authors: Chi-Fu Hsiao, 蕭吉富
• Other Authors: Ming-Hwa Sheu
• Format: Others
• Language: zh-TW
• Published: 2002
• Online Access: http://ndltd.ncl.edu.tw/handle/40791988496465428717

碩士 === 國立雲林科技大學 === 電子與資訊工程研究所碩士班 === 90 === In this paper, we use the color coordinate transformation and the contain JND discrete wavelet transform to build the prototype of JPEG2000 system. In order to speed up advance the non-linear color transformation, in the part of non-linear color transformation, we propose a new fast and efficient cubic-root algorithm and hardware architecture in accordance with non-linear transformation XYZ L*a*b*. It not only promotes the transform speed but also has reconstruction quality equal to the way of traditional transform. In the way of DWT, we follow the JPEG2000 standard that tile the original image into many block sizes, because every time there is a portion of image data will be processed, this way is more flexible compare to that the whole original image is handled in the traditional, and it significantly reduces the memory during processing. Whereas in the JND quantization aspect, we make use of the human visual sensitivity that has different spatial frequency and orientation response in different sub-band, we give it different quantization threshold weight, compare to the exclude JND characteristics quantization in the way of convention, it has better reconstruction quality and higher compression ratio. In the investigation of wavelet filter through the computer simulation, we found an excellent quality performance group among a lot of wavelet filters which also be able to adopt both of the luminance and chromaticity components. As far as the color transformation and contain JND DWT benchmark, we did an advanced study on the efficiency of the image transform/coding and new hardware architecture designation in order to enable the color transformation and contain JND characteristic DWT to be more closely in practicality for the future digital commercial applications.