In recent years, with the rapid growth of the Internet as well as the increasing demand for broadband services, live pay-television broadcasting via the Internet has become a promising business. To get this implemented, it is necessary to protect distributed contents from illegal copying and redistributing after they are accessed. Fingerprinting system is a useful tool for it. This paper shows that the anti-collusion code has advantages over other existing fingerprinting codes in terms of efficiency and effectivity for live pay-television broadcasting. Next, this paper presents how to achieve efficient and effective anti-collusion codes based on unital and affine plane, which are two known examples of balanced incomplete block design (BIBD). Meanwhile, performance evaluations of anti-collusion codes generated from unital and affine plane are conducted. Their practical explicit constructions are given last.

A high speed 3D shape reconstruction method with multiple video cameras and multiple computers on LAN is presented. The video cameras are set to surround the real 3D space where people exist. Reconstructed 3D space is displayed in voxel format and users can see the space from any viewpoint with a VR viewer. We implemented a prototype system that can work out the 3D reconstruction with the speed of 10.55 fps in 313 ms delay.

This paper presents an algorithm for string searching in a character lattice. A character lattice, which is obtained through a character recognition process, is a general and flexible data structure that represents many hypothesized strings in a document image. In this paper, the authors propose a simple and efficient algorithm; it consists of a single loop of some set-operations and scans the character lattice only once. The authors also describe two actual implementations of the algorithm; one uses Bit-Arrays and the other a Trie. Owing to its bir parallelism, the Bit-Array approach is able to search for a single pattern faster than the Trie approach, and is easily extended to complex matchings such as an approximate one. It is suited for document retrieval systems that need to search for a keyword as fast as possible. A hashed compact version of the character lattice is also useful to increase the speed of the search for a single pattern. In contrast, the Trie approach is able to search for a large number of patterns simultaneously, and is suited for document understanding systems that need to extract words from the character lattice. The experimental results have shown that both approaches achieve high performance.

Blind deconvolution (BD) is the problem of restoring sharp images from blurry images when convolution kernels are unknown. While it has a wide range of applications and has been extensively studied, traditional shift-invariant (SI) BD focuses on uniform blur caused by kernels that do not spatially vary. However, real blur caused by factors such as motion and defocus is often nonuniform and thus beyond the ability of SI BD. Although specialized methods exist for nonuniform blur, they can only handle specific blur types. Consequently, the applicability of BD for general blur remains limited. This paper proposes a shift-variant (SV) BD method that models nonuniform blur using a field of kernels that assigns a local kernel to each pixel, thereby allowing pixelwise variation. This concept is realized as a Bayesian model that involves SV convolution with the field of kernels and smoothing of the field for regularization. A variational-Bayesian inference algorithm is derived to jointly estimate a sharp latent image and a field of kernels from a blurry observed image. Owing to the flexibility of the field-of-kernels model, the proposed method can deal with a wider range of blur than previous approaches. Experiments using images with nonuniform blur demonstrate the effectiveness of the proposed SV BD method in comparison with previous SI and SV approaches.

Multiple-shot person re-identification, which is valuable for application in visual surveillance, tackles the problem of building the correspondence between images of the same person from different cameras. It is challenging because of the large within-class variations due to the changeable body appearance and environment and the small between-class differences arising from the possibly similar body shape and clothes style. A novel method named “Bi-level Relative Information Analysis” is proposed in this paper for the issue by treating it as a set-based ranking problem. It creatively designs a relative dissimilarity using set-level neighborhood information, called “Set-level Common-Near-Neighbor Modeling”, complementary to the sample-level relative feature “Third-Party Collaborative Representation” which has recently been proven to be quite effective for multiple-shot person re-identification. Experiments implemented on several public benchmark datasets show significant improvements over state-of-the-art methods.

The amount of seasonings used during food preparation is quite important information for modern people to enable them to cook delicious dishes as well as to take care for their health. In this paper, we propose a near real-time automated system for measuring and recording the amount of seasonings used during food preparation. Our proposed system is equipped with two devices: electronic scales and a camera. Seasoning bottles are basically placed on the electronic scales in the proposed system, and the scales continually measure the total weight of the bottles placed on them. When a chef uses a certain seasoning, he/she first picks up the bottle containing it from the scales, then adds the seasoning to a dish, and then returns the bottle to the scales. In this process, the chef's picking and returning actions are monitored by the camera. The consumed amount of each seasoning is calculated as the difference in weight between before and after it is used. We evaluated the performance of the proposed system with experiments in 301 trials in actual food preparation performed by seven participants. The results revealed that our system successfully measured the consumption of seasonings in 60.1% of all the trials.

Vignetting is a common type of image degradation that makes peripheral parts of an image darker than the central part. Single-image devignetting aims to remove undesirable vignetting from an image without resorting to calibration, thereby providing high-quality images required for a wide range of applications. Previous studies into single-image devignetting have focused on the estimation of vignetting functions under the assumption that degradation other than vignetting is negligible. However, noise in real-world observations remains unremoved after inversion of vignetting, and prevents stable estimation of vignetting functions, thereby resulting in low quality of restored images. In this paper, we introduce a methodology of image restoration based on variational Bayes (VB) to devignetting, aiming at high-quality devignetting in the presence of noise. Through VB inference, we jointly estimate a vignetting function and a latent image free from both vignetting and noise, using a general image prior for noise removal. Compared with state-of-the-art methods, the proposed VB approach to single-image devignetting maintains effectiveness in the presence of noise, as we demonstrate experimentally.

The Hough voting framework is a popular approach to parts based pedestrian detection. It works by allowing image features to vote for the positions and scales of pedestrians within a test image. Each vote is cast independently from other votes, which allows for strong occlusion robustness. However this approach can produce false pedestrian detections by accumulating votes inconsistent with each other, especially in cluttered scenes such as typical street scenes. This work aims to reduce the sensibility to clutter in the Hough voting framework. Our idea is to use object segmentation and object pose parameters to enforce votes' consistency both at training and testing time. Specifically, we use segmentation and pose parameters to guide the learning of a pedestrian model able to cast mutually consistent votes. At test time, each candidate detection's support votes are looked upon from a segmentation and pose viewpoints to measure their level of agreement. We show that this measure provides an efficient way to discriminate between true and false detections. We tested our method on four challenging pedestrian datasets. Our method shows clear improvements over the original Hough based detectors and performs on par with recent enhanced Hough based detectors. In addition, our method can perform segmentation and pose estimation as byproducts of the detection process.

Re-identifying the same person in different images is a distinct challenge for visual surveillance systems. Building an accurate correspondence between highly variable images requires a suitable dissimilarity measure. To date, most existing measures have used adapted distance based on a learned metric. Unfortunately, real-world human image data, which tends to show large intra-class variations and small inter-class differences, continues to prevent these measures from achieving satisfactory re-identification performance. Recognizing neighboring distribution can provide additional useful information to help tackle the deviation of the to-be-measured samples, we propose a novel dissimilarity measure from the neighborhood-wise relative information perspective, which can deliver the effectiveness of those well-distributed samples to the badly-distributed samples to make intra-class dissimilarities smaller than inter-class dissimilarities, in a learned discriminative space. The effectiveness of this method is demonstrated by explanation and experimentation.

A method for extracting and recognizing character strings from unformed document images, which have inclined character strings and have no structure at all, is described. To process such kinds of unformed documents, previous schemes, which are intended only to deal with documents containing nothing but horizontal or vertical strings of characters, do not work well. Our method is based on the idea that the processes of recognition and extraction of character patterns should operate together, and on the characteristic that the character patterns are located close to each other when they belong to the same string. The method has been implemented and applied to several images. The experimental results show the robustness of our method.