In this paper, a new scheme for fast video retrieval is proposed. In the scheme, a video is represented by a set of feature vectors which are computed using the robust alpha-trimmed average color histogram. To efficiently retrieve videos, the locality sensitive hashing technique, which involves a uniform distance shrinking projection, is applied. Such a technique does not suffer from the notorious "curse of dimensionality" problem in handling high-dimensional data point sets and guarantees that geometrically close vectors are hashed to the same bucket with high probability. In addition, unlike the conventional techniques, the involved similarity measure incorporates the temporal order of video sequences. The experimental results demonstrate that the proposed scheme outperforms the conventional approaches in accuracy and efficiency.

In this paper, a new method for document image watermarking is proposed. The method first divides the document into weight-invariant partitions in the spatial domain using a novel image feature - the neighborhood pixel ratio. Such a ratio effectively summarizes the local information and is robust to common attacks. The document image is then partitioned in a key-dependant way for high capacity and security. Our experimental results demonstrate the effectiveness of the new method.

A novel general paradigm for text detection using a support vector machine (SVM) is proposed. Unlike prevailing techniques in the literature, our adaptive SVM incorporates information from each input image. In addition, for better classification results and higher efficiency, a novel kernel called the Fre/spl acute/chet kernel is presented for SVM classification. The adaptive SVM aims to serve as a general paradigm to improve the prevailing techniques. In the experiment, we apply the paradigm to a simple algorithm and successfully obtain a new competitive method for text detection.

In this paper, we propose a novel image watermarking algorithm, which is based on the scheme where the watermark is embedded into projection of an image onto the secret set of key-dependant basis functions. We propose a key-dependant image decomposition method, which involves solving a bottleneck hamiltonian path problem by an ant colony optimization technique. Combined with the decomposition technique, the original scheme is significantly improved on the aspects of speed, capacity and security. In particular, our algorithm usually shortens the running time of the original scheme by a factor of several hundred.