A Semi-automatic Alternative Text Generation System For Web Image Accessibility

Today, as the society is based on highly developed information technology, the disabilities still meet great trouble in web accessibility. The disabilities, especially the ones with vision disability, cannot enjoy the efficient and convenient brought by the Internet. Nowadays, web accessibility correction technology aims at helping the disability, by making them have a convenient access to the inaccessibility web pages. Exiting web accessibility correction will send parts, which do not meet the requirement of accessibility guidelines, directly to volunteers who work for the web accessibility project. However, this procedure will take a lot of manual work as well as bring personal view to the annotation of web images.For the people with vision disability, this paper proposes a new semi-automatic alternative text generation algorithm for web images correction. With image salient detection and image annotation techniques, the method can generate alternative text candidate set for inaccessibility webs images. These candidate sets can be delivered to crowdsourcing platform further, for the use of auxiliary in web image annotation and reconstruction work. Compared with the existing web images accessibility correction method, this method can make the volunteers of web accessibility pay more attention in main content of image. This method can also help them provide more neutral annotation for web images. Taking more advantage of saliency detection results, the method can also provide optional alternative text template. Experiments show that the proposed method has a certain practical performance, and will be a great helpful in semi-automatic web images correction system...