We propose an advanced motion-compensated prediction method for improving the coding efficiency in H.264/AVC. Nine separable two-dimensional interpolation filters are applied to precise compensation for motion in various directions. A new optimal cost function, which considers the bit rate and distortion for coding the macroblock, is also proposed. The filter is adaptively selected per macroblock to minimize the proposed cost function. Also, an algorithm for reducing the overhead of transmitting the filter coefficients is described. In experimental results on the various standard QCIF/CIF test video sequences, the proposed method shows improvement in coding efficiency over conventional methods. This leads to approximately a 7.30% (for one reference frame) and 4.46% (for five reference frames) bit-rate reduction on average, compared to H.264/AVC.
We propose a new efficient discrete cosine transform (DCT) and quantization algorithm for the H.264 encoder. We theoretically analyze integer DCT and quantization of the H.264 encoder. A theoretical analysis shows that there are redundant calculations for DCT and quantization. In the proposed algorithm, one of four methods—1. DCT skip; 2. reduced DCT1; 3. reduced DCT2; 4. original DCT—is selected by using the minimum SAD information obtained in the motion estimation. The reduced DCT1 and reduced DCT2 eliminate the redundant calculations required for DCT and quantization. The simulation results show that the proposed algorithm achieves an approximately 15 to 25% computational saving without video-quality degradation, compared to the conventional method.
In this paper, we describe the method for color reproduction based on the spectral reflectance. The accuracy of color calibration depends on the number and the distribution of the color patches as well as color calibration method. The effect of dot overlap between the neighboring dots makes the printed color patches significantly darker. Therefore we propose a new design of the color patches where the distribution of the printed spectral reflectances is uniform in the spectral domain. We measure the spectral reflectances of the color patches with an inkjet printer. The basis functions are extracted from the measured patch data using principal component analysis by Karhunen-Loeve expansion. A neural network is used to transform the coefficients of principal component analysis to CMY colorants. This patch design increases the accuracy of the conversion of the neural network. The accuracy of color reproduction is evaluated according to the number and the distribution of the color patches in terms of the root mean square error and the color difference.
This paper describes a new model based error diffusion method to compensate dot overlapping. We designed 32 test patterns to measure printer non-linearity. The effects of four neighboring pixels are considered in the standard error diffusion. The effect of dot overlap is exactly measured with the conditions of various distribution of ON/OFF neighboring pixels. A new model based error diffusion based on measured non-linearity shows good reproduction of gray scale. It is worth while to notice that there is little assumption about the radius of dot, homogeneity inside the dot, and overlapped area phenomenon. The proposed method is applicable to real environment with little restrictions of the printer dot modeling.
In this paper, a novel and unified hardware structure to implement various binarization algorithms is proposed. It is designed to perform: 1) simple thresholding, 2) high pass filtering, 3) dithering, 4) blue noise masking, 5) error diffusion, 6) threshold nodulated error diffusion, and 7) edge enhanced error diffusion. In general, these algorithms have been implemented with several logic blocks. We found that a single data path architecture can be used for implementing those algorithms. A new structure is designed to have same data-flow that can share the blocks. All processing is possible in the proposed unified architecture which is based on the threshold modulated and edge enhanced error diffusion scheme. This structure has error filter coefficient registers, error memory, threshold memory, and arithmetic units, etc. This paper shows that the proposed hardware structure reduces the number of gates efficiently. The hardware design and debugging complexity is reduced by the unified control logic and data path.
Dithering with blue noise mask has no artifact as the error diffusion and it is very simple to implement. But it does not represent well the local property of edges. In this paper, a local adaptive masking technique is proposed. By controlling the local adaptive factors, the amount of high pass filtered and blue noise components are controlled. It was shown that the proposed algorithm has similar effects on simple threshold in background, blue noise masking in mid- level and high pass filtering in edge regions. The resulting image shows good gray-tone reproduction as well as good edge characteristic.
Wavelet transform is used efficiently for high compression ratio image coding. It is useful for a high resolution and color document image processing system. However, a large amount of memory is required in wavelet decompression for a large image. Conventional wavelet transform does not allow reconstructing a sub-region of the image. Therefore, it is hard to use wavelet transform in a color document image processing system. In this paper, a block wavelet transform method is proposed. An image is divided into blocks and the samples of the adjacent block are used to transform one block for removing the edge artifact. The required number of samples of the adjacent block are derived. By using the proposed method, the memory requirement for wavelet coding/decoding is reduced to that of the block size. Any targeted region of an image can be compressed or reconstructed. The proposed method can be used for a color document image processing system.
In this paper, a new edge enhanced error diffusion algorithm which is based on the Eschbach's algorithm is proposed. Thick edge artifacts with large edge enhancing factor as well as less edge enhancement effects for the bright or dark pixel values are observed in the previous algorithm. By analyzing the phenomena, a new improved algorithm is proposed by using the diffused error sum and input pixel value. An input pixel is classified into a normal- or edge-region pixel based on the error sum criteria. Then, a new error calculation is employed for the edge-region pixel, while conventional error calculation is used for the normal- region pixel. The proposed method requires only a few additional calculations, and provides edge enhanced binary output images. The edge is less influenced by the brightness offset and thick edge artifacts are reduces.
new edge-enhanced error diffusion algorithm, based on Eschbach's algorithm, is proposed. Thick-edged artifacts as well as small edge-enhancement effects for the bright or dark pixel values are observed in the previous algorithm. By analyzing the phenomena,
a new improved algorithm is proposed by using the diffused error sum and input pixel value. An input pixel is classified into a normal- or edge-region pixel based on the error sum criterion. A new
error calculation is then employed for the edge region pixel, while conventional error calculation is used for the normal-region pixel. The proposed method requires only a few additional calculations and provides edge-enhanced binary output images. The edges are in fluenced less by the brightness offset, and thick-edged artifacts are reduced.
In this paper, we propose a new image compression technique using wavelet transform and human visually estimated noise sensitivities. These consist of frequency, background brightness, and edge height sensitivities. The background brightness sensitivity for each quantizing point is modeled by a quadratic function. The edge sensitivity for each quantizing point is modeled by a non-linear function. The minimum value becomes background brightness sensitivity and edge height sensitivity is multiplied by the frequency sensitivity for determining the quantization step size. Quantization step sizes are calculated by using coefficients of lowest frequency band which are coded losslessly. Therefore, in the proposed method, information to specify quantization step size for higher frequency band, is not needed. The coefficients of high frequency bands are arithmetically coded in horizontal and vertical directions depending on the edge direction. Compared with previous human visual systems based image compression methods, the proposed method shows improved image quality for the same compression ratio with less computational cost.
In the general fractal image compression, each range block is approximated by a contractive transform of the matching domain block under the mean squared error criterion. In this paper, we propose a fractal image compression algorithm with perceptual distortion measure rather than the mean squared error. In the perceptual distortion measure, the background brightness sensitivity and edge sensitivity are used. To obtain the sensitivity of the background brightness for each pixel, the average value of the neighborhoods is calculated and applied to a quadratic function. In the edge sensitivity for each pixel, sum of the differences in the neighborhood is calculated and applied to a nonlinear function. The perceptual distortion measure is obtained by the multiplications of the background brightness sensitivity, the edge sensitivity, and the error between the range block and the transformed domain block. For the range blocks having large distortion, they are splitted and the same algorithm is applied for smaller blocks. Compared to the method with the mean squared error measure, 10% compression ratio improvement under the same image quality is achieved.