【摘要】：視覺系統(tǒng)是人類最為重要的感知器官。經(jīng)研究發(fā)現(xiàn),大部分盲人都只是在視覺通路上部分發(fā)生病變,這就為視覺假體的研究奠定了生理基礎(chǔ)。視覺假體技術(shù)屬于功能點(diǎn)刺激技術(shù)的一種,通過在盲人眼中植入微電極陣列來刺激其視覺通路中未受損傷的部分,產(chǎn)生光幻視。但目前視覺假體技術(shù)受低分辨率、光幻視的不穩(wěn)定性以及光幻視延時等問題的限制,這就需要通過圖像預(yù)處理在少量的像素點(diǎn)下實(shí)時為盲人提供識別物體的主要信息。邊緣是圖像信息最集中的部分,盲人通過對邊緣圖像的識別能保證基本的視覺感受。而小波分析是近年來逐漸興起的熱門信號處理方法,其良好的時頻局部分析的特性以及多尺度多分辨率思想在圖像處理中具有高度的應(yīng)用價值。本文在小波變換的理論基礎(chǔ)上設(shè)計了基于提升小波的邊緣檢測算法。重點(diǎn)設(shè)計了基于FPGA的提升小波變換VLSI電路結(jié)構(gòu),提出了一種9輸入的并行計算電路,保證了少量像素點(diǎn)的邊緣連續(xù)性,并在每級變換的高頻運(yùn)算中加入閾值截取,提取出更為精細(xì)的邊緣曲線。論文中首先在MATLAB軟件上設(shè)計算法并對各類邊緣提取算法的效果進(jìn)行對比分析,之后詳細(xì)設(shè)計介紹了算法的硬件電路并對算法各個模塊進(jìn)行RTL級設(shè)計及仿真,最后通過CMOS攝像頭、SDRAM存儲器、VGA顯示器等搭建視頻處理系統(tǒng)來對算法進(jìn)行系統(tǒng)級驗證。本文設(shè)計的提升小波運(yùn)算屬于整數(shù)域的運(yùn)算,具有同址計算、相比傳統(tǒng)小波速度更快、易重構(gòu)等適合硬件實(shí)現(xiàn)的特性,并在視頻流處理的過程中圍繞SDRAM進(jìn)行分時復(fù)用,在提取出較為準(zhǔn)確的邊緣信息的同時能夠滿足視覺假體所需求的分辨率大小的視頻實(shí)時傳輸,豐富了視覺假體圖像處理的研究內(nèi)容。
[Abstract]:Visual system is the most important organ of human perception. It is found that most blind people are only partially affected in the visual pathway, which lays a physiological foundation for the study of visual prosthesis. Visual prosthesis is a kind of function point stimulation technology, which stimulates the undamaged parts of visual pathway by implanting microelectrode array in blind eyes, which produces optical illusion. However, the current visual prosthesis technology is limited by the problems of low resolution, optical illusion instability and optical illusion delay, which need to provide the blind with the main information of object recognition in real time under a small number of pixels through image preprocessing. Edge is the most concentrated part of image information. Blind people can guarantee basic visual feeling by recognizing edge image. Wavelet analysis is a popular signal processing method in recent years. Its good characteristics of time-frequency local analysis and multi-scale and multi-resolution theory have high application value in image processing. Based on the theory of wavelet transform, an edge detection algorithm based on lifting wavelet is designed in this paper. The VLSI circuit structure of lifting wavelet transform based on FPGA is mainly designed. A 9-input parallel computing circuit is proposed, which ensures the edge continuity of a small number of pixels, and the threshold is added to the high frequency operation of each level transform. Finer edge curves are extracted. Firstly, the algorithm is designed on MATLAB software, and the effect of all kinds of edge detection algorithms is compared and analyzed. Then, the hardware circuit of the algorithm is introduced in detail, and each module of the algorithm is designed and simulated at RTL level. Finally, the video processing system is built by CMOS camera, SDRAM memory and VGA display to verify the algorithm at system level. The lifting wavelet operation designed in this paper belongs to the integer domain operation, has the same address calculation, is faster than the traditional wavelet, easy to reconstruct and so on suitable for the hardware realization, and carries on the time-sharing multiplexing around the SDRAM in the video stream processing process. While extracting more accurate edge information, it can meet the need of visual prosthesis resolution of real-time video transmission, enrich the research content of visual prosthesis image processing.
【學(xué)位授予單位】：西安理工大學(xué)
【學(xué)位級別】：碩士
【學(xué)位授予年份】：2016
【分類號】：TP391.41

【參考文獻(xiàn)】

相關(guān)期刊論文 前7條

1 劉紫燕;;實(shí)時圖像邊緣檢測的設(shè)計及FPGA實(shí)現(xiàn)[J];電子科技;2011年12期

2 任秋實(shí);;視覺假體的研究進(jìn)展與面臨的挑戰(zhàn)[J];生命科學(xué);2009年02期

3 祝強(qiáng);徐瑾;張鐸;張建華;;小波模極大值邊緣檢測算法研究[J];武漢理工大學(xué)學(xué)報(信息與管理工程版);2008年06期

4 孫蕾;羅建書;;小波變換點(diǎn)對稱邊界延拓問題研究[J];計算機(jī)應(yīng)用;2008年02期

5 劉云清;佟首峰;姜會林;;利用FPGA實(shí)現(xiàn)SDRAM控制器的設(shè)計[J];長春理工大學(xué)學(xué)報;2005年04期

6 田志明,楊軍,羅嵐;異步復(fù)位設(shè)計中的亞穩(wěn)態(tài)問題及其解決方案[J];電子器件;2002年04期

7 趙家良;“視覺2020”行動與我國防盲治盲工作[J];中華眼科雜志;2002年10期

相關(guān)碩士學(xué)位論文 前2條

1 鮑書朋;人工視覺假體系統(tǒng)設(shè)計[D];鄭州大學(xué);2014年

2 解程程;視覺假體圖像處理策略以及編碼研究[D];上海交通大學(xué);2011年

，

本文編號：2169905