WebInSight: Making Web Images Accessible
Jeffrey P. Bigham, Ryan S. Kaminsky, Richard E. Ladner,
Oscar M. Danielsson and Gordon L. Hempton
Abstract
Images without alternative text are a barrier to equal web access for blind users. To illustrate the problem, we conducted a series of studies that conclusively show that a large fraction of significant images have no alternative text. To ameliorate this problem, we introduce WebInSight, a system that automatically creates and inserts alternative text into web pages on-the-y. To formulate alternative text for images, we present three labeling modules based on web context analysis, enhanced optical character recognition (OCR) and human labeling. The system caches alternative text in a local database and can add new labels seamlessly after a web page is downloaded, resulting in minimal impact to the browsing experience.

Keywords
Web accessibility, web studies, transformation proxy, optical character recognition

1. INTRODUCTION
Blind users do not currently have equal access to the web. Images are used in navigation bars, as form buttons and to display textual and visual content, but, unless web authors provide alternative text for these images, blind users employing screen readers and refreshable Braille displays are left to guess the images’ contents. In our studies we found that a large fraction of images lack alternative text. For example, of the significant images found on the homepages of the 500 most high-traffic websites[1], only 39.6% were as signed alternative text.
Illustrating the problem, the homepage of the UCLA Computer Science Department (Figure 1) contains 30 images that should have alternative text, but only two (6.67%) were assigned any text at all. As a result, a blind user may have difficulty navigating this page. As another example, the University of Michigan Computer Science and Engineering Department provides a listing of all faculty on its website along with their contact information (Figure 2). Unfortunately, each e-mail address is presented as an image without equivalent alternative text, presumably in a misguided attempt to thwart e-mail harvesters1. The unintended consequence, however, is that blind users who want to e-mail these professors either cannot do so or must find their e-mail addresses through other means. The problems highlighted here are characteristic of many sites throughout the web.
WebInSight ：制作網(wǎng)頁(yè)圖片的訪問(wèn)
Jeffrey P. Bigham, Ryan S. Kaminsky, Richard E. Ladner,
Oscar M. Danielsson and Gordon L. Hempton
摘要
文盲可使用圖象進(jìn)行網(wǎng)頁(yè)瀏覽。為了說(shuō)明問(wèn)題，我們進(jìn)行了一系列的研究，得出結(jié)論表明，很大一部分重要的圖像沒(méi)有別的選擇文本。為了改善這個(gè)問(wèn)題，我們引進(jìn)WebInSight ，系統(tǒng)會(huì)自動(dòng)創(chuàng)建并插入替代文字到網(wǎng)頁(yè)上的信號(hào)。制定替代文字的圖片，我們提出三個(gè)模塊貼標(biāo)基于Web的背景分析，增強(qiáng)文字標(biāo)識(shí)符識(shí)別（ OCR ）和人力標(biāo)記。該系統(tǒng)緩存替代文字在當(dāng)?shù)氐臄?shù)據(jù)庫(kù)，并可以添加新的標(biāo)簽后，無(wú)縫的網(wǎng)頁(yè)下載，從而影響最小的瀏覽體驗(yàn)。

關(guān)鍵詞
網(wǎng)頁(yè)可讀性，網(wǎng)絡(luò)的研究，轉(zhuǎn)化代理，文字標(biāo)識(shí)符識(shí)別

1.導(dǎo)言
文盲目前沒(méi)有平等地使用網(wǎng)頁(yè)。圖片中使用的導(dǎo)航欄，作為形式按鈕和顯示的文字和視覺(jué)的內(nèi)容，但除非網(wǎng)絡(luò)作者提供替代文字這些圖片，文盲使用屏幕閱讀器和refreshable盲文顯示器左猜的圖像內(nèi)容。在我們的研究，我們發(fā)現(xiàn)，很大一部分缺乏圖像替代文字。舉例來(lái)說(shuō)，重要的圖像在網(wǎng)站上的500最高流量的網(wǎng)站[ 1 ] ，只有39.6 ％ ，分別為替代文字。
深入這個(gè)問(wèn)題，加州大學(xué)洛杉磯分校計(jì)算機(jī)科學(xué)系（圖1 ）包含30個(gè)影像，應(yīng)該有替代圖片，但只有兩個(gè)（ 6.67 ％ ）被指派任何文字占全部。因此，一個(gè)盲人的用戶可能會(huì)遇到困難瀏覽此網(wǎng)頁(yè)。另一個(gè)例子是，美國(guó)密歇根大學(xué)計(jì)算機(jī)科學(xué)與工程系提供了一個(gè)清單，把所有教師在其網(wǎng)站上與他們的聯(lián)系信息（圖2 ） 。不幸的是，每個(gè)e-mail地址是作為一個(gè)形象不等于替代文字，大概是在誤導(dǎo)試圖阻撓電郵harvesters1 。意想不到的后果，然而，是誰(shuí)盲目用戶想用e - mail這些教授也不能這樣做或必須找到自己的e - mail地址通過(guò)其他手段。這些問(wèn)題突出的特點(diǎn)是這里的許多地盤(pán)整個(gè)網(wǎng)絡(luò)。

圖像，要么有一個(gè)行動(dòng)與他們有聯(lián)系的（例如，鏈接或按鈕）或彩色的和大于一定規(guī)模的特別關(guān)注。當(dāng)這些圖像沒(méi)有替代文字，可無(wú)障礙嚴(yán)重減少。 WebInSight系統(tǒng)的介紹了目標(biāo)，例如signicant圖像和提供了一種機(jī)制，自動(dòng)插入適當(dāng)?shù)奶娲淖?。它通過(guò)處理網(wǎng)絡(luò)請(qǐng)求，并改變了返回網(wǎng)頁(yè)上飛行。作為轉(zhuǎn)變過(guò)程中，三個(gè)新的坐標(biāo)，形象標(biāo)識(shí)模塊針對(duì)這一領(lǐng)域，利用方法的基礎(chǔ)上加強(qiáng)網(wǎng)絡(luò)方面的分析，光學(xué)字符識(shí)別nition （光學(xué)字符識(shí)別）和人力標(biāo)記。
在本文中，我們提出了一系列的網(wǎng)絡(luò)研究，證明所觀察到的問(wèn)題，討論的架構(gòu)和實(shí)施WebInSight ，并nally ，描述標(biāo)記模塊用于指定替代文字任意圖像在..