Katsutoshi Sawada

Scalable coding schemes based on DCT and MC prediction

The paper describes scalable coding schemes which use DCT and motion compensated interframe prediction. In the scalable coding, a lower resolution picture can be obtained by decoding only a partial bitstream, while a full resolution picture is obtained by decoding the total bitstream. Two types of scalable coding scheme are studied. In the first type (schemes A), an input picture is first decomposed by DCT, then MC prediction coding is carried out in the DCT picture domain. In the second type (scheme B), MC prediction is first carried out in the full resolution picture and then DCT decomposition is performed for the prediction difference picture. The coding performance of these two schemes was estimated by computer simulation experiments. A performance comparison between scalable and non-scalable coding schemes was also carried out. The experimental results have demonstrated that scheme B is superior to schemes A.

Fractal coding of color images using the correlation between Y and C components

This paper presents an efficient fractal coding scheme for color images and demonstrates its experimental results. The proposed fractal coding scheme utilizes the correlation between a luminance component (Y) and two color difference components (Cr and Cb) of an input color image. The Y, Cr and Cb components are first decomposed to low and high frequency sub-band images. Fractal block coding is performed only on the lowest frequency sub-band images of Y, Cr and Cb. The other high frequency sub-band images are encoded by vector quantization (VQ). In the fractal coding process for Y, each range block is encoded by a set of contractive affine transformations of its correspondent domain block. For Cr and Cb, on the other hand, only the range block average values are coded. The other fractal coded data of the correspondent range block of Y are applied also to Cr and Cb. The computer simulation experimental results show that the coded and decoded color images obtained by the proposed scheme give higher SNR values and better image qualities compared to the conventional fractal coding scheme and JPEG.

Subband-based scalable coding schemes with motion-compensated prediction

This paper describes scalable coding schemes which use subband picture decomposition and motion compensated interframe prediction. In the scalable coding, a lower resolution picture can be obtained by decoding only a subset of the total bitstream, while a full resolution picture is obtained by decoding the total bitstream. Two types of scalable coding schemes are studied. In the first type (schemes A), an input picture is first decomposed to subband pictures, then MC prediction coding is carried out in the subband picture domain. In the second type (scheme B), MC prediction is first carried out in the full band picture domain and then subband decomposition is performed for the prediction difference picture. Coding performance for these two types of schemes was estimated by computer simulation experiments. The performance comparison between scalable and non scalable coding schemes was also carried out. The experimental results have demonstrated that the scheme B is superior to schemes A.

Fast global motion estimation algorithm based on elementary motion detectors

This paper presents a fast global motion estimation algorithm based on so called elementary motion detectors or EMD. EMD, modeling insect visual signal processing systems, have low computational complexity aspects and thus can be key components to realize such a fast global motion estimation algorithm. The developed algorithm is evaluated by being applied to various types of image sequences and is found to provide accurate estimation results.

Real Time Parameter Optimization for Elementary Motion Detectors

Elementary motion detectors are biologically motivated motion detectors based on a visual processing model of insects. Due to their low computational costs in estimating motion vectors, they can realize real time motion detection algorithms. In order to obtain good estimation results, however, it is necessary to set up parameters configuring these detectors properly. This paper proposes a real time algorithm to optimize such parameters. The algorithm has been implemented in a hardware system with an FGPA device and a CMOS camera module. It has been found that parameters can be dynamically adapted in real time to improve estimation results.

RLGS profile segmentation via a SVM

This paper presents a new and fast segmentation algorithm, capable of incorporating experts knowledge and expertise in a straightforward manner, to detect DNA spots in a restriction landmark genomic scanning (RLGS) profile. A RLGS profile is a gray scale image obtained in a X-ray film containing hundreds of dark spots corresponding to certain DNA fragments labeled by radioactive markers. Extracting these DNA spots is a very challenging task due to their fuzzy boundaries against noisy background regions having non-uniform intensity variations. The proposed algorithm, utilizing a support vector machine, is found to be successful to detect and extract DNA spots in a RLGS profile.

A Scalable Video Coding Scheme Based on Adaptive Infield/Inframe DCT and Adaptive Frame Interpolation

Publisher Summary This chapter describes a spatio-temporal resolution scalable coding scheme. Resolution scalability means a coding property where lower partial resolution pictures can be obtained by decoding only subsets of the total coded bit stream, while the full resolution picture is reconstructed by decoding the total bit stream. This scheme employs frame sub-sampling associated with adaptive interpolation for temporal scalability and adaptive infield/inframe discrete cosine transform (DCT) for spatial scalability. The proposed scheme provides four different spatio-temporal resolutions of a video sequence—two temporal resolutions, each consisting of two spatial resolutions. This can be applied to interlaced video sequences effectively. Computer simulation results have demonstrated that this scheme has a better coding performance compared to conventional non adaptive methods.

Entropy-constrained quantization for existing video decoders

Once video coding standards are finalized, decoders are completely specified so that encoders from different providers can maintain the inter-operability. However, some of the encoding parameters such as the decision thresholds of the embedded quantizers and the buffer-control schemes still remain at designers freedom. Taking advantage of this freedom, this paper concentrates on the design of encoders for existing decoders. More specifically, an entropy- constrained design approach is described only for the quantizer decision thresholds within encoders while the reconstruction levels and the variable-length code (VLC) table remain unchanged. The efficiency of the new method is demonstrated through an example of the well- known Lloyd-Max quantizers operating on broad-tailed generalized Gaussian distributed memoryless sources.

Fast local motion estimation algorithm using elementary motion detectors

This paper presnts a fast local motion estimation algorithm based on so called elementary motion detectors or EMDs. EMDs, modeling insect’s visual signal processing systems, have low computational complexity aspects and can thus be key components to realize such a fast local motion estimation algorithm. The contribution of the presented work is to introduce dual parameter estimators or DPEs by configuring EMDs so that they can estimate local motions in terms of both direction and speed mode parameters simultaneously. The estimated local motion vectors are displayed as arrows superimposed over video image frames. The developed algorithm is implmented in a DirectShow application by using Mircosoft’s DirectX runtime library and is evaluated using various types of video image sequences. It is found to be able to estimate local motion vectors in real time even in moderate PC computing platforms and hece no high profile hardware devices are needed for its real time operation.

DNA spot extraction via grayscale range quantization and morphological measurement

This paper presents a DNA spot extraction algorithm by grayscale range quantization and measuring morphological features or a shape index of DNA spot candidate regions. Due to fuzzy boundaries of DNA spots and intensity variations of background regions in RLGS profile images, extracting DNA spots is a very challenging task. The developed algorithm first obtains DNA spot candidate regions by quantizing the intensity values of a RLGS profile image. The following elimination process of false DNA spots is done by computing and thresholding the shape index.