摘要
空间时间延迟积分电荷耦合器件(TDICCD)相机图像压缩2维(2-D)离散小波变换模块的基于常用硬件实现方法实现效率低,为此提出一种适于电荷耦合器件(CCD)图像的(128,N)快速2-D提升小波实现结构.在CCD图像行方向上,采用16个提出的1维(1-D)行提升小波变换模块快速实现结构并行分解图像,每个模块采用4个延时寄存器的值作为中间值并用2步预测更新步骤融合的结构来计算行小波系数;在行小波系数的列方向上,提出了一种基于多路复用技术的(32,16)列提升小波变换结构,最终实现了2-D提升小波变换.结果表明,2-D提升小波变换模块能快速稳定地工作,分解1帧图像仅需60.198μs,与传统方法相比,节省时间达到49.90%,节省逻辑资源10.71%,节省寄存器资源11.80%,节省存储器资源12.98%,有效地解决了CCD图像压缩小波变换硬件实现效率低的问题.
In order to resolve low efficiency of hardware implementation of 2-dimension(2-D) discrete wavelet transform(DWT) based on the method for space time delay and integration charge coupled devices(TDICCD) compression system,a fast implementation structure(128,N) of 2-D lifting-based DWT is proposed.In the direction of line of CCD image,16 modules of the fast implementation structure of 1-dimension(1-D) line lifting-based DWT are used to decompose image in parallel.Each module uses the value of four delay registers as the intermediate value to compute line wavelet coefficient,which is computed by a fusion structure of two step of prediction and updating.In the direction of line wavelet coefficient,a(32,16) column lifting-based DWT structure based on multiplexing technology is proposed,which finally completes 2-D lifting-based DWT.The experiment results show that a module of 2-D lifting-based DWT can work fast and stably.The time needed in decomposing a frame image is only 60.198 μs.Compared with traditional approaches,the time consuming is decreased by 49.90%,logical resource is decreased by 10.71%,register is decreased by 11.80%,and memory is decreased by 12.98%.Low efficiency of hardware implementation of wavelet module of CCD image compression in space camera is effectively improved.
引文
[1]Hsu Chun lung,Huang Ming da,Chun-Lung Hsu,et al.Design of an error-tolerance scheme for discrete wavelettransform in JPEG2000 encoder[J].IEEE Transactionson Computer,2011,60(5):628-637.
[2]Fernando García-Vílche,Joan Serra-Sagristà.Extendingthe CCSDS recommendation for image data compressionfor remote sensing scenarios[J].IEEE Transactions onGeosciences and Remote Sensing,2009,47(10):3431-3445.
[3]Stefan Strahl,Heiko Hansen,Alfred Mertins.A dynamicfine-grain scalable compression scheme with application toprogressive audio coding[J].IEEE Transactions on Au-dio,Speech,and Language Processing,2011,19(1):14-23.
[4]Wladimir J van der Laan,Andrei C Jalba,Jos BTM Ro-erdink.Accelerating wavelet lifting on graphics hardwareusing CUDA[J].IEEE Transactions on Parallel and Dis-tributed Systems,2011,22(1):132-146.
[5]Stephan Rein,Martin Reisslein.Low-memory wavelettransforms for wireless sensor networks:a tutorial[J].IEEE Communications Surveys&Tutorials,2011,13(2):291-307.
[6]Chao Cheng,Parhi K K.High-speed VLSI implementa-tion of 2-D discrete wavelet tansform[J].IEEE Transac-tions on Signal Processing,2008,56(1):393-403.
[7]Mohanty B K,Pramod Kumar Meher.Memeory efficientmodular VLSI architecture for high throughput and low-la-tency implementation of multilevel lifting 2-D DWT[J].IEEE Transactions on Signal Processing,2011,59(5):2072-2084.
[8]Jiang Zhe,Gao Zhiyu,Liu Yanyan,et al.An efficientVLSI architecture for lifting-based discrete wavelet trans-form[J].IEEE Transactions on Circuits and System-Ⅱ:Express Briefs,2012,59(3):158-162.