In subband image coding, each subband signal is encoded by appropriate quantizer after an input image is divided into some subbands. According to the stochastic characteristics of input image, the optimum frequency band partition which can minimize the quantization noise power in the reconstructed image was proposed previously, in order to improve the coding performance for subband image coding. However, the partition pattern on the 2-D frequency domain of the previous method was invariable when the bit rate is changed. This paper presents new optimum frequency band partition which can determine both the optimum partition on the 2-D frequency domain and the optimum combination of quantizers with subband signals for a given bit rate. In the proposed method, the calculation algorithm with lower computational cost to obtain the optimum frequency band partition has been developed, in paying attention to the kurtosis of the quantized subband signals. It is seen in the experimental results that the improvement of the proposed method is about 0.5[dB] at the lower bit rate compared with the previous method.

View full abstract

We propose an adaptive pyramidal approach using local feature for orientation code matching (OCM). OCM can achieve robust image matching in outdoor environment, and its high computational costs can be reduced by using pyramidal approach. However, as the pyramid levels deepen, the mismatch between pyramid generation method and types of scene tends to cause the decline of matching success rate. To avoid this problem, we broadly classify the image features into two types: artificial features or not. And we define a new degree of artificial structure based on local orientation code distribution. In this way, we obtained an adaptive selection of pyramid generation method according to the local image features. Our experimental results using aerial images showed that our method worked well and improved the success rate at the upper pyramid levels.

View full abstract

The privacy of unknown person may be included in the digital images unintentionally taken as a sight. The action of the file upload of it would occur invasion of his/her privacy. The JPEG files may include Exif information which is regarded as personal privacy. For these two privacy problems, we propose two privacy protection techniques for JPEG files using reversible information hiding. One is to be able to protect personal privacy if we embed an Exif information into the privacy region. The other is to be able to protect personal privacy when we erase Exif information from the JPEG format and degrade the privacy region by adding some noise. In addition, we embed Exif information and the Huffman code of DCT coefficients in the privacy region into a whole image. In the experiments, we show an example of privacy protection and investigate the feasibility of our method in the real world.

View full abstract

A conditional image enhancement for the low-quality document image is proposed. In a document recognition by software such as OCR, image resolution is very important factor for recognition rate and an images which are low-resolution can be expected to be improved through the existing enhancement methods which can add high-frequency component. However, if images include such noise regions as bleed-through, blot and slur, they process not only improving character regions but also increasing a noises, and cause false recognition. The proposed method can improve the character regions without noise enhancement by conditional adding high-frequency component. This paper shows the effective using OCR software and by comparing existing enhancement method.

View full abstract

This paper proposes a method for estimating extrinsic camera parameters using video images and sparsely acquired GPS data. Although the extended bundle adjustment methods which fuse vision and GPS data have been already proposed and they work finely if GPS data are accurately acquired, existing methods have the following problems: (1) The accuracy of estimated camera position largely depends on the confidence of GPS positioning data because they do not consider the GPS positioning confidence.
(2) Solution often converges to local minima when GPS data can not be acquired for long time. In order to solve these problems, we add weighting coefficients depending on the GPS positioning confidence to an energy function for extended bundle adjustment. In order to avoid local minima, estimated camera positions without GPS data are fitted to GPS positions before optimization.

View full abstract

The methods which use many cameras or Omni-Directional sensors to acquire a 3D free viewpoint image are well known. Another method to generate the free viewpoint image uses a distance mapping. The problem is that the former is complicated and expensive system, and the latter caused an occlusion. In order to solve these problems, we proposed a system which used one camera and one spin mirror to generate the free viewpoint 3D image. However, in the generated free viewpoint image there is a problem to which some edges become discontinuous. In this paper, we propose a method to generate the smooth 3D free viewpoint image by scaling interpolation. The image sequences are scaled in the lateral direction to change gradient of an edge in image sequences. If the gradient of the edge in image sequences changes, the endpoint position of the edge in image sequences will move up and down. So this method can connect the endpoints of an edge between adjoining two image sequences and generate a smooth free viewpoint image. The validity of the proposal method was verified by experiment.

View full abstract

In the time-shared reconstructing process, the quality of the reconstructed images depend severely on the setting conditions of the dividing process to the partial holograms of the input object. In fact, in the process reported last year (conventional process), since the skipping span of the points of the partial holograms were not always taken constant, the images were not reconstructed with uniform brightness. Therefore, in order to reduce uneven brightness in the time-shared reconstruction, it is important to study the characteristics of the reconstructed images corresponding to the setting conditions of the partial holograms and the point interval of the input objects. In this paper, we investigate the brightness in the images recovered by time-sharing technique corresponding to the setting conditions on the number of the partial holograms and the point interval. First, we show the results obtained from evaluation of the reconstructed images by changing the point interval. Next, based on the obtained results, we show the results obtained by the time-shared multiplex reconstructing process to the input object with multiple describing points.

View full abstract

We propose a simple binary arithmetic coder STT-coder, in which arithmetic coding operation can be executed by referring a state transition table. At first as a fundamental study, we introduce the idea of STT-coder for 3-bit interval register, and evaluate its coding performance. The coding efficiency is found to be satisfactory despite of simplification of the coding process, in which renormalization is done only when the MSB of the current address is determined. Then, to improve the performance for high MPS probability sources, we extend the interval register size to 6 bits, and study the optimization of the specific parameter called as “offset,” which shows the difference of the bottom address of the current valid interval and the address given by the already determined code bit stream. The optimal number of the kind of offset values and its combination are introduced and coding performance under the optimal design is evaluated.

View full abstract

We have proposed STT-coder; a simple binary arithmetic coder, in which encoding and decoding can be executed by referring a state transition table, and found that STT-coder having a 6-bit interval register with 8 probability states for the MPS probability range of 0.5 to 0.987 achieves satisfactory static coding performance for practical application. In this paper, we examined the probability estimation method based on a state transition rule, and newly introduced a concept of “LPS transit ratio” to especially improve the coding efficiency of the sources having a high LPS probability, and found that the introduction of LPS transit ratio achieves 2% improvement of the coding efficiency for high LPS probability sources and a flat performance for the covered area. Further 4% improvement can be realized by increasing the number of probability states to 16, which is almost 1.5% better coding efficiency than that of MQ-coder, a conventional arithmetic coder, having an equivalent number of states for the MPS range covered by the STT-coder .

View full abstract

Recently, biometric authentication is widely used, therefore robbery and fear of property loss automatically has been reduced. However, biometric has its own disadvantages for examples; falsification and external factor. In this paper, we are using tooth for the biometric authentication in order to solve the problems. Acquisition of the tooth is done through sanitary method for non-contact type, which is different from fingerprint and vein verification. Generally, the main application of the identification using the tooth is for person confirmation of the dead body. X-ray photography and treatment mark of the tooth are used for identification with the help of special machine and tools. In this work also, the dentition image is measured by CCD camera, where the individual identification systems use both general and local features of the tooth. The experimental results of this system have shown the effectiveness of the proposed system. It is confirmed that the false rejection rate, FRR=0% and false acceptance rate, FAR=0%.

View full abstract

We propose a non-photorealistic rendering method for generating snow-noise-like color images from color photographs. Snow-noise-like color images are generated with iterative enhancement bilateral filter which performs the iterative calculation of bilateral filter and unsharp mask. In order to evaluate the performance of the proposed method for generating snow-noise-like color images, we conduct experiments using some example images.

View full abstract