外文文獻(xiàn)及翻譯-fpga實(shí)現(xiàn)實(shí)時(shí)適應(yīng)圖像閾值-資料下載頁(yè)
2025-05-12 12:19本頁(yè)面
【導(dǎo)讀】硬件架構(gòu)是基于一種加權(quán)聚類算法的架構(gòu),這。該方法采用聚類的二值加權(quán)神經(jīng)網(wǎng)絡(luò)法找到。兩個(gè)像素組的質(zhì)心。更新是基于輸入像素的灰度級(jí)和相關(guān)權(quán)值的差額的，通過學(xué)習(xí)快慢因素來衡量其速率。硬件系統(tǒng)是在FPGA平臺(tái)上實(shí)現(xiàn)的，它包含兩個(gè)功能模塊。第一個(gè)模塊獲得圖像框架閾值，另一。兩個(gè)模塊的并行性和簡(jiǎn)單的硬件組成部分使其適用于實(shí)時(shí)應(yīng)用程序，并且，其性能可與經(jīng)常用于離線閾值技術(shù)相媲美。通過利用FPGA對(duì)無數(shù)的例子進(jìn)行模擬和實(shí)驗(yàn)，得到。這項(xiàng)工作的基本應(yīng)用是確定激光的質(zhì)心，但接下來將會(huì)討論它在其他方面的應(yīng)用。圖像二值化是圖像處理的一個(gè)主要問題。背景色的灰度級(jí)是不同的。現(xiàn)在已有一些較好的使圖像二值化地算法，就性能而不是就速度而言，這。在激光測(cè)距，即確定目標(biāo)的運(yùn)動(dòng)范圍的過程中，所捕獲的圖像為二值圖像。描并處理大約超過100頁(yè)的文件。由于將圖像轉(zhuǎn)換為二值圖像，可以在不。硬件對(duì)數(shù)實(shí)現(xiàn)和標(biāo)準(zhǔn)偏差計(jì)算使這
【正文】 criteria for thresholding performance 7 evaluation in two different contexts, document images and NDT (nondestructive testing) images. It employed an average of five performance criteria: misclassification error, edge mismatch, relative foreground area error, modified Hausdorff distance, and region nonuniformity. All top ranked methods belong to clustering and entropybased thresholding categories. The ranking also considers the subjective evaluation on the visual outlines of the extracted object. From the hardware implementation point of view, the effectiveness of a thresholding method may also be considered in terms of other parameters such as speed and plexity. These bee very important in realtime image processing applications. All of the highranked cluster based techniques have to pute some image features, such as the histogram, maximum/minimum gray level values, or variance of image, before calculating the threshold value. Therefore an image must be preprocessed pixel by pixel. For these methods a large processing overhead is present. In the entropybased techniques plex putational processes, such as logarithms, are also required. {In hardware implementation logarithm and standard deviation calculation makes the hardware requirement for these methods plicated.} Moreover, the methods require considerable processing time after the full image is available to pute the threshold. Although these discussed methods have good performance, they are not generally suitable for realtime implementation. Alternatively we can enhance or modify these techniques. The basic requirement for the thresholding method is its adaptability and efficiency. It should also have the least dependency on image preprocessing. 3 PROPOSED APPROACH Clusteringbased methods are one the high ranked thresholding techniques [1]. In this method, gray level pixels of an image are divided into two clusters, foreground and background. There are several approaches for clustering a set of input gray level pixels. Artificial neural works (ANN) are especially useful for classification and clustering problems. Talukdar and Sridhar [18] used an artificial neural work structure as a clustering technique, called weightedbased clustering threshold (WCT). The weightedbased clustering method uses the clustering property of artificial neural works to calculate a threshold, where the threshold is average of the centroids of these two clusters. Artificial neural works are the simple clustering of the primitive nodes. This clustering occurs by creating layers, which are then connected to one another. The processing ability of the work is stored in the interlayer connection strengths, called weights, which are obtained by a process of learning from a set of training patterns. Each of inputs to the node is multiplied by a connection weight. During recent decades, very diverse categories of ANN have been introduced by researchers. Each category is applicable to a specific domain and proposing a general neural work to solve all problems seems to be impossible. One of the proposed solutions, which is applicable to the classification and image segmentation problems, is Unsupervised Competitive Learning [19].
畢業(yè)設(shè)計(jì)相關(guān)推薦
鄂ICP備17016276號(hào)-1