Search

This paper presents a method for the dehazing of images through the application of the atmosphere model associated with McCartney Optics. The proposed method is based on the assumption that similarities among hazy blocks are associated with similarities in transmissions. A coarse transmission map was fabricated using vector quantization and then refined using a guided image filter. Subtracting the effects of atmospheric light from hazy images makes it possible to obtain haze-free images. Experiment results demonstrate that the proposed method produces images that appear more natural than those obtained using other methods.

In real-time video transmission, the loss of packet results in visual quality degraded for the succeeding frames. This paper proposes an efficient full frame algorithm for depth-based 3-D videos. Each frame can be regarded as combination of objects. True motion estimation (TME) and depth map are exploited to calculate the motion vector (MV) for each object. In this paper, the object is defined as the pixels with the same MV and similar depth value. The object-based MV can extrapolate each object in reference frame to reconstruct the damage frame. In the consideration of computational complexity, in this method, only the high frequency regions need to execute TME. In this paper, we provide a new method to obtain the objects according to temporal and depth information. From the simulation results, our algorithm gives better visual quality and PSNR in most case with lower complexity.

Following the advances in single-sensor imaging techniques, interest in producing a zoomed full-color image from a Bayer mosaic data has been increased. Almost all of the recent approaches identified, with respect to the demosaicking step in the imaging pipeline, have chiefly focused on misguidance problems. However, in regions consisting of sharp edges or fine textures, these approaches are prone to large blurring effects. This paper proposes a new joint solution to overcome the above problems associated with demosaicking and zooming operations. On the basis of an enhanced soft-decision framework, we estimate the edge features by computing the integrated gradients. This allows the extraction of gradient information from both color intensity and color difference domains, simultaneously. Then, the edge guidance is incorporated in the interpolation of various stages to preserve edge consistency and improve computational efficiency. In addition, an edge-adaptive, iterative, back-projection technique is developed to compensate for image blurring as well as to further suppress color artifacts. Experimental results indicate that the new algorithm produces outstanding objective performances and sharp, visually pleasing color outputs, when compared to numerous other single-sensor image zooming solutions.

A shadow-removal algorithm based on cycle generative adversarial network is proposed. There are two networks in the proposed method. To increase the diversity of shadow images for improving the robustness in the shadow-removal process, the first network is used to add shadows in nonshadow images to increase the variation of training data. Then, the second network is trained for the shadow-removal task. Six different losses are calculated and combined in the loss function to increase the performance. Ablation experiments show that the resulting images suffer from some artifacts without any of the six losses in the loss function. The proposed method presents lower value of root-mean-squared error and the superior visual quality compared to the state-of-the-art image removal algorithms.

The application of image super-resolution technologies in recent years has increased noticeably. The main purpose of image up-scaling is to obtain high-resolution images from low-resolution images, and these up-scaled images should keep satisfactory visual qualities and present natural textures. The most popular image up-scaling algorithms are based on interpolation methods in spatial domain. However, the up-scaled images may produce blurring artifacts. Therefore, using spatial sharpening filters is usually used to make blurred images sharp and clear. The quantity of image sharpening is the key to decide the visual qualities of up-scaled images. In this paper, a method based on self-similarity of images and using simple linear regression to build a reconstruction model for improving visual qualities of up-scaled images adaptively is proposed. The experimental results show that our algorithm provides better subjective visual qualities as well as the peak signal-to-noise ratio (PSNR).

When photographs are being taken in an outdoor environment, the medium in air will cause light attenuation and further reduce image quality, and this impact is especially obvious in a hazy environment. Reduction of image quality results in the loss of information, which renders an image recognition system unable to identify objects in the image. In order to eliminate the hazy effect on images and improve the visual quality, this paper presents an efficient method combining the fuzzy inference system and the neural network filter to solve image dehazing. During dehazing, the fuzzy inference system is adopted to estimate the variations in light attenuation, and the erosion of morphological operation and the neural network filter are used to eliminate the halation and achieve optimization in transmission map refinement. Finally, the brightest 1% of the atmospheric light is utilized to calculate the color vector of atmospheric light to eliminate color cast. The experimental results indicate that the proposed method is superior to other dehazing methods.

Along with digitizing and multimedia era, the image has not changed from the original entity into any changes can be dealt with digital preservation methods. Although the digital image capture technology means more and more developed, but there are still many variables affect the quality of an image. An image quality usually depends on the user's usage or changes in the natural environment. Due to the natural environment of the most common factors that influence is light, so an image of the brightness distribution over the target object caused by extreme hardly recognizable condition common. Therefore, we will use the independent component analysis of an input color images Red, Green, and Blue three Color Space to the main component analysis, in order to achieve the target tracking and analysis.

This paper presents a robust true motion estimation algorithm, designated as MPMVP (Multi-pass and Motion Vector Propagation), to enhance the accuracy of the motion vector fields in frame rate up-conversion applications. The MPMVP uses a multi-pass scheme to progressively refine approximate motion vectors to true motion vectors based upon the motion information acquired in previous passes. The multi-pass motion estimation process uses a large block size to detect the motion vectors within the objects themselves and small block sizes to detect the motion vectors along the object boundaries. Actually, the block size is progressively reduced during the search process. When the motion vector of a block is considered to be sufficiently accurate for motion estimation purposes, the block is said to be converged and the local motion vector search process terminates. A novel technique, referred to as motion vector propagation, is then applied to propagate the motion vector of the converged block to its neighboring blocks. This technique not only ensures that neighboring motion vectors within the same object have a high degree of spatial correlation, but also accelerates the convergence of the motion vectors in the neighboring blocks and therefore reduces the overall computational time and expense of the multi-pass motion vector search procedure. A novel distortion criterion is proposed to enhance the tolerance of the traditional sum-of-absolute-difference measurement technique applied in the motion estimation scheme to noise and shadow effects. The experimental results demonstrate that the proposed true motion estimation algorithm outperforms the traditional full search, 3DRS and TCSBP algorithms in terms of both the smoothness of the generated motion vector fields and the visual quality of the up-converted frames.

A good fast motion search algorithm should efficiently speed up the encoding time and keep the quality of encoded video stable at the same time. Researches have shown that many fast algorithms lose the quality requirement in some special video sequences. These video sequences often have heavy motions and need large search windows for motion vector search. E3SS, DS, and E-HEXBS, which are famous algorithms, are not good enough in these sequences. As to UMHexagonS, it is able to meet the high video quality requirement very well, but it costs too much computation. This paper introduces a multi-stage motion estimation algorithm. The algorithm ensures getting good video quality while decreases the motion search time efficiently. It divides the search regions into many un-overlapped small-diamond regions and forces the motion search to go outward for larger motion vectors. This method is also designed to avoid mistaking local optimal motion vectors. For this reason, the selected motion vector is refined by several stages. Experimental results show that the proposed algorithm uses almost the same number of checking points as E3SS but achieves a better quality. Furthermore, the proposed algorithm is also tested in H.264/AVC JM9.5 encoder; the experimental results show that this algorithm is also suitable for variable block-size motion estimation.

In this paper, a near-duplicate video retrieval method developed based on invariant features was proposed. After shot change detection, Zernike moments are extracted from each key-frame of videos as invariant features. We obtain the key-frame similarity by computing the difference of Zernike moments between key-frames of the query and test videos. To achieve near-duplicate video retrieval, each key-frame is considered as an individual sensor and then evaluating all of key-frames is considered as multiple sensors. The results of key-frames are fused to obtain a better performance of near-duplicate video retrieval. The experimental results show that the proposed method can not only find the relevant videos effectively but also resist to the possible modifications such as re-scaling and logo insertion.

An image watermarking model based on visual secret sharing (VSS) is proposed for protection of ownership. In the embedding phase, the watermark is first divided by VSS into two parts: a hiding watermark and a secret watermark. Then only the hiding watermark is embedded into the original image and the secret watermark is reserved for watermark extracting by the owner. In the extracting phase, the hiding watermark is extracted from the watermarked image first and then directly superimposed on the secret watermark to recover the watermark information.

Light absorption and scattering in underwater environments can result in low-contrast images with a distinct color cast. This paper proposes a systematic framework for the enhancement of underwater images. Light transmission is estimated using the random forest algorithm. RGB values, luminance, color difference, blurriness, and the dark channel are treated as features in training and estimation. Transmission is calculated using an ensemble machine learning algorithm to deal with a variety of conditions encountered in underwater environments. A color compensation and contrast enhancement algorithm based on depth information was also developed with the aim of improving the visual quality of underwater images. Experimental results demonstrate that the proposed scheme outperforms existing methods with regard to subjective visual quality as well as objective measurements.

Contrast enhancement algorithms have been designed to conform the contrast to human visual perception. Histogram Equalization (HE) is widely used for contrast enhancement, but the problem is over-enhancement, washed out, and detail loss for some parts of the processed image when the discontinuous histogram. Modified histogram equalization based on the partitioned dynamic range was proposed in this paper which to solve the problems of HE. We propose a method which combines the Probability Density Function (PDF) and the Range Distribution Function (RDF). PDF aimed at keeping the characteristic of the original image and the RDF aimed for reducing the devastating effect of the equalization.

This paper presents a novel algorithm for speeding up the codebook design in image vector quantization that exploits the correlation among the pixels in an image block to compress the computational complexity of calculating the squared Euclidean distortion measures, and uses the similarity between the codevectors in the consecutive code-books during the iterative clustering-process to reduce the number of codevectors necessary to be checked for one codebook search. Verified test results have shown that the proposed algorithm can provide almost 98% reduction of the execution time when compared to the conventional Linde-Buzo-Gray (LBG) algorithm.

This paper describes a new single bit-map block truncation coding (SBBTC) scheme that works with a Hopfield neural network (HNN) for the coding of color images. An incoming color block is encoded using three block-truncation coding (BTC) encoders with a common bit map for each of the three primary color planes. An HNN is used to generate a good single bit map in the SBBTC by clustering the color vectors in the block into two classes. The considering block is then encoded by the generated bit map and the two reconstruction colors each associated to one of the two classes. In comparison with other existing methods, the proposed color BTC gives the best mean-square error (MSE) performance at the same bit rate.

With the improvement in technology of thin-film-transistor liquid-crystal display (TFT-LCD), the resolution requirement of display becomes higher and higher. Super-resolution algorithms are used to enlarge original low-resolution (LR) images to meet the visual quality of the high-resolution (HR) display. In this research, an efficient super resolution algorithm is proposed. The proposed algorithm consists of two steps. First, the Lanczos interpolation is used for LR images to get the preliminary HR images. For solving the over-smoothing problems generally caused by interpolation, it needs to add texture information to refine the preliminary HR images. Subsequently, a refinement process based on simple linear regression and the self-similarity between a pair of LR and HR images is performed to provide proper information of textures. In the experimental results, the proposed algorithm not only performs well in the objective measurement such as PSNR, but also in visual qualities.

The main objective of image enhancement is to improve the visual quality of digital images that are captured under extremely low or non-uniform lighting conditions. We present an adaptive image enhancement algorithm based on Zone System. This study reveals hidden image details and increases the contrast of an image with low dynamic range. It is comprised two processes: adaptive luminance enhancement and adaptive contrast enhancement. The adaptive luminance enhancement algorithm is a global intensity transform function based on Zone System information. This process not only increases the luminance of darker pixels but also compresses the dynamic range of the image. The adaptive contrast enhancement adjusts the intensity of each pixel based on the discontinuities of the local luminance. It also improves the contrast of local region and reveals the details of image clearly. The proposed algorithm has good performance on enhancing contrast, preserving more detail of characteristics and sharpening edges of objects in experimental results. The performance with our proposed was better evaluation and comparison than other algorithms in the subjective and objective evaluation.

A new method for facial expression recognition is presented in this paper. Our method offers basically three benefits: first, it is combining regional compression with wavelet decomposition to preprocess the facial image, which can effectively remove interference and enhance useful information. Second, local directional number pattern (LDN) is adopted to extract the facial feature, it is not only encode the intensity transitions of the image but also encode the directional information, which is more effective to represent the feature. The last, sparse representation classification method is used to represent the test expression image by the linear combination of the training expression images. Results show that our method is effective.

In this paper, we propose a passive scheme for photorealistic computer generated image (PRCG) identification and device brand classification. To this end, a periodic phenomenon of variance of pixel values resulting from demosaicing process is analyzed in RGB color space. Based on the phenomenon, the proposed scheme is composed of color space transformation, high-pass filtering, feature extraction, and classification. To make the phenomenon obvious, a high-pass filter is performed. The periodic phenomenons in different directions are measured as features in the Fourier domain and combined with a hierarchical classifier for PRCG detection and brand identification simultaneously. Experimental results show that our proposed scheme can not only detect photographic images (PIM) but also determine what brand of a camera is used.

The Consultative Committee for Space Data Systems proposed an efficient image compression standard that can do lossless compression (CCSDS-ICS). CCSDS-ICS is the most widely utilized standard for satellite communications. However, the original CCSDS-ICS is weak in terms of error resilience with even a single incorrect bit possibly causing numerous missing pixels. A restoration algorithm based on the neighborhood similar pixel interpolator is proposed to fill in missing pixels. The linear regression model is used to generate the reference image from other panchromatic or multispectral images. Furthermore, an adaptive search window is utilized to sieve out similar pixels from the pixels in the search region defined in the neighborhood similar pixel interpolator. The experimental results show that the proposed methods are capable of reconstructing missing regions with good visual quality.

In this paper, we present an improved multi-pass true motion estimation algorithm for frame rate up-conversion. In the proposed motion estimation algorithm, the motion vectors of different objects which are parted with depth information will be refined respectively. Therefore, more accurate pixeliVbased motion vector can be obtained which causes the performance of Frame Rate Up-Conversion (FRUC). For overlapped regions, the motion vector can be chosen in candidates with depth information. Experimental results show that our proposed algorithm effectively enhances the overall quality of the frame rate up-converted video sequence, both subjectively and objectively, even with complex scenes and fast motion.

Satellite images have been more and more popularly used in our daily life. For satellite image compression, the Consultative Committee for Space Data Systems (CCSDS) had proposed an image compression standard (CCSDS-ICS) which is so far the most widely implemented in hardware. CCSDS-ICS provides good compression performance under the measure of PSNR. However, PSNR is a concept of average differences between the original image and the reconstructed one, which can not guarantee that the maximum difference of the two images would be less than an expected error. For using the information of satellite images more precisely without modifying the existing hardware system of CCSDS-ICS, a fearlessness strip-based satellite image compression algorithm is proposed. An adaptive pre-quantizer in order to reduce the entropy of an image is adopted and the quantized data is then fed into CCSDS-ICS as an input to be processed in the loss less compression mode. Experiment results show that by the proposed algorithm, not only can the good performance under the measure of PSNR be preserved, but also the differences are limited in the expected range to increase the reliability and the precision of the use of compressed satellite images.

Due to the development of multimedia hardware and computer devices, single-image super resolution has recently become a hot topic. The main concept of super resolution is to obtain a high-resolution image from a low-resolution one, and the former should look like it had been acquired with a sensor having the resolution of the up scaled image or, at least, present a i¨naturali¨ texture. Most existing single-image super resolution techniques upscale a low-resolution image by either spatial-domain based or wavelet-domain based algorithms. Edge information in the spatial domain can be detected and enhanced to construct a sharp high-resolution image while in the wavelet domain, the support of filters to model the regularity of natural images is exploited. The proposed algorithm combines the advantages of both spatial-domain based and wavelet-domain based algorithms where the back-projection technique is adopted to minimize the reconstruction error with an efficient iterative procedure. Comparing with conventional image interpolation techniques, simulation results show that the proposed algorithm is considerably superior in both objective and subjective terms. Finally, for lowering down the hardware cost, some improvement is presented that can properly reduce the scan lines according to the limit of the input buffer.

This paper proposes a method for facial expression recognition. Facial feature vectors are generated from keypoint descriptors using Speeded-Up Robust Features. Each facial feature vector is then normalized and next the probability density function descriptor is generated. The distance between two probability density function descriptors is calculated using Kullback Leibler divergence. Mathematical equation is employed to select certain practicable probability density function descriptors for each grid, which are used as the initial classification. Subsequently, the corresponding weight of the class for each grid is determined using a weighted majority voting classifier. The class with the largest weight is output as the recognition result. The proposed method shows excellent performance when applied to the Japanese Female Facial Expression database.

Goal of proposed algorithm is to enhance images like the digital camera photos or, better. Some enhancement filters process too excessive that makes image look unnatural. The paper proposes an algorithm to represent obviously the detailed textures and sharpen the overall image based on perceptual approach. Then the sharpened images are merged by the edge information of original image. The experimental results show that the detailed textures can be easily observed. The whole image looks not only more sharp but also natural after applying our algorithm.

In real-time video transmission, the loss of packet results in visual quality degraded for the succeeding frames. This paper proposes an efficient full frame algorithm for depth-based 3-D videos. Each frame can be regarded as combination of objects. True motion estimation (TME) and depth map are exploited to calculate the motion vector (MV) for each object. In this paper, the object is defined as the pixels with the same MV and similar depth value. Each object in reference frame can be extrapolated to reconstruct the damage frame by the object-based MV. In the consideration of computational complexity, in this method, only the high frequency regions need to execute TME. From the simulation results, our algorithm gives better visual quality and PSNR in most case.

The ultimate goal of realistic graphics is the creation of images that provoke the same responses that a viewer would have to a real scene. In the field of computer graphics, one related key problem is the issue of displaying radiant intensities in a meaningful way. The paper proposed an efficient local contrast enhancement algorithm by histogram equalization and histogram stretching method. The algorithm also combines the tone mapping to fix high Contrast Scene Images. Experimental results have shown that it produces low dynamic range images with perceptually good quality and strong detail preservation. These strong experimental results warrant further exploration of the potential of this algorithm.

JPEG standard is commonly used for the still image compression. However, the DCT-based coding of JPEG is one of the lossy compression tools and introduces most artifacts in the decompressed image such as blocking and ringing artifacts. In this paper, our proposed method focuses on how to efficiently detect the blocks which cause ringing artifacts. Moreover, these ringing blocks will be further filtered and the texture region will be able to preserve as well as the smooth region. Simulation results show that our proposed method can outperform the related algorithms in subjectively and objectively. By comparing JPEG compressed image, the decompressed image using our algorithm can achieve the better PSNR as well as the visual performance, especially at the lower quality coding (the higher compression rate).

In packet-based video transmissions, packets loss may result in the loss of the whole video frame due to channel errors. A lot of error concealment algorithms have been proposed to overcome the channel errors, but most of them only can deal with the loss of macro blocks but whole frame. In this thesis, we propose an efficient full frame recovery algorithm which is based on object-based module for a new coding standard H.264/AVC. First, our proposed method estimates the true motion of each block in the reference frame. Besides, the reference frame will divided into several objects according to the estimated motion vectors. Furthermore, each object will be extrapolated to its corresponding place in the missing frame. Finally, the experimental results show that our algorithm could provide a more efficient solution to recover a missing whole frame than previous methods. In addition, based on the object-based module, our proposed method provides the better visual quality than other comparative algorithms.

In the modern compression system, besides high compression qualities and low bit rates, others extra functional capabilities become more and more important. For the use of satellite images, the data size is always very large and the users might be just interested in some regions of a satellite image. Thus it is an ideal way to keep the better quality in the regions of interest than in others in a satellite image. In JPEG2000 standard part I, the region of interest (ROI) is implemented by MAXSHIFT, which needs to generate an ROI mask with extra memory space. For the computers in a satellite, lower computing complexity and fewer memory are required. Thus the Consultative Committee for Space Data Systems Recommendation (CCSDS) proposes a specification of image data compression for satellite images. An ROI extension based on CCSDS is proposed in our study. Experimental results show that it saves more compression time and moreover it is competitive with the JPEG2000 standard without the memory used for masks.

Ultrasound images are the important bases of disease diagnostic. Unfortunately, the qualities of ultrasound images are generally limited due to speckle noises. Speckle reduction is an important pre-processing step in the ultrasound image feature extraction, analysis and recognition. In this paper, we present an approach for ultrasound image enhancement. It is designed to utilize the three technologies: the separability and multiresolution properties of the wavelet, the local statistics of each subband and the edge enhancement of the shape. The performance of the proposed method has been compared with that of the commonly novel approaches on both synthetic speckle images and real medical ultrasound images. The proposed method reveals superior performance in term of the PSNR value and perceptible quality. Because of the superior performance in noise reduction and edge preservation, the proposed method is more suitable than the other methods in computer-aided diagnosis.

This paper proposes asystem forthe facial expression recognition. Firstly, we perform noise reduction by a median filter of facial expression image. Then, a cross-correlation of optical flow and mathematical models from the facial points are used. To define these facial points of interest in the first frame of an input face sequence image, which utilize manually marker. The facial points were automatically tracked by a cross-correlation, which is based on optical flow,and then extracted the feature vectors. The mathematical model extracts features from the feature vectors. An ELMAN neural network was applied to classify expressions. The performances of the proposed facial expressions recognition were computed by Cohn---Kanade facial expressions database. This proposed approach achieved a high recognition rate.

Digital Cameras that gradually replace conventional cameras store photographs in digital format. As a result, captured images can be more easily processed. Consequently the quality of image is more important. In this paper, we present an overview of image quality attributes of different contrast enhancement in gray level images. Moreover, we present results of analysis for some contrast enhancement methods. We also present the evaluation of existing contrast enhancement methods with regard to these attributes.

In this paper, we proposed a method to enhance global contrast of an image based on Average Luminance with Weighted Histogram Equalization (ALWHE) for Low Dynamic Range Images (LDRI). Automation of parameter selection can easily be done by iteratively using the proposed method. The purposed method overcomes the disadvantage of HE like luminance shifting and washed-out looking. We present experimental results to compare results of contrast enhancement with other methods.

We present a simple and fast algorithm for image noise estimation. The input image is assumed to be corrupted by additive zero mean Gaussian noise. To exclude structures or details from contributing to the noise variance estimation, a simple edge detection algorithm using first-order gradients is applied first. Then a Laplacian operator followed by an averaging over the whole image will provide very accurate noise variance estimation. There is only one parameter which is self-determined and adaptive to the image contents. Simulation results show that the proposed algorithm performs well for different types of images over a large range of noise variances. Performance comparisons against other approaches are also provided.

Studies of contrast sensitivity of the human eyes show that logarithmic curves obey the Weber-Fechner law of just noticeable difference response in human perception. In this paper, we propose a local contrast enhancement algorithm with logarithm-based curves for both high dynamic and low dynamic images and this algorithm can adaptively change the curvature with local information. We also define two parameters to decide the level of contrast enhancement in the tone mapping procedure. For halo artifact which is suffered from local operator, a two-stage procedure is designed to solve the problem. We present experimental results to show performance of our algorithm compared with other existing methods.

The attenuation of the light transmitted through air can reduce image quality when taking a photograph outdoors, especially in a hazy environment. Hazy images often lack sufficient information for image recognition systems to operate effectively. In order to solve the aforementioned problems, this study proposes a hybrid method combining fuzzy theory with weighted estimation for the removal of haze from images. A transmission map is first created based on fuzzy theory. According to the transmission map, the proposed method automatically finds the possible atmospheric lights and refines the atmospheric lights by mixing these candidates. Weighted estimation is then employed to generate a refined transmission map, which removes the halo artifact from around the sharp edges. Experimental results demonstrate the superiority of the proposed method over existing methods with regard to contrast, color depth, and the elimination of halo artifacts.

The goal of image segmentation is to partition the image into several homogenous regions or particular objects with similar properties, such like color and texture. It is essential for object recognition and image retrieval, and it also could enhance the performance of certain compression algorithms. In this paper, a perceptual color contrast based watershed algorithm for color images segmentation is presented. The conventional watershed segmentation algorithm is not suitable for color image directly, since it is hard to decide a suitable color gradient measurement for color image. The proposed algorithm adopts HSV color space to define the color contrast gradient for the improvement of watershed algorithm. Experimental result shows that the proposed algorithm can avoid the over segmentation problem, and produce segmented result which is close to human perception.

Discrete cosine transform (DCT) is widely used in many practical image/video compression systems because of its compression performance and computational efficiency. This work adopts DCT, and modified the SPIHT algorithm that designed initially for encoding the discrete wavelet transform (DWT) coefficients in order to suit to encode DCT coefficients. The algorithm represents the DCT coefficients to concentrate signal energy and proposes combination and dictator to eliminate the correlation in the same level subband for encoding the DCT-based images. The coding complexity of the proposed algorithm for DCT coefficients is just close to JPEG but the performance is higher than JPEG2000. Furthermore, the proposed algorithm append to deblocking function in low bit rate in order to improve the perceptual quality. Experimental results indicate that the proposed technique improves the quality of the reconstructed image in terms of both PSNR and the perceptual results over JPEG2000 at the same bit rate.

In the paper, an 8/spl times/8 DCT approach is adopted to perform subband decomposition, followed by modified SPIHT data organization and entropy coding. The translation function has the ability to retain the detail characteristics of an image. By means of a simple transformation to gather the DCT spectrum data with the same frequency domain, the translation function exploits all the characteristics of all individual blocks to a global framework. In this scheme, insignificant DCT coefficients that correspond to the same spatial location in the high-frequency subbands can be used to reduce the redundancy by a combined function proposed in associated with the modified SPIHT. Simulation results showed that the embedded DCT-CSPIHT image compression reduced the computational complexity to only a quarter of the wavelet based subband decomposition and improved the quality of the reconstructed medical image in terms of both the peak PSNR and the perceptual results over JPEG2000 and the original SPIHT at the same bit rate.

Various medical image compression techniques have been proposed for accelerating image propagation in many applications. JPEG2000 is a new generation technique that can encode near lossless ultrasound images at medium bit-rate with diagnostically acceptable quality. Because the coder of JPEG2000 is based on wavelet transform, the reconstructed image will contain some ringing artifacts. Some de-ringing algorithm must be applied to enhance image quality. This study presents quad-tree decomposition and a set of morphological filters for reducing the ringing artifacts of ultrasound images. Specifically, the presented morphological filters use eight predefined morphological operations, including four structuring elements (SE) that include both dilation and erosion. The proposed voting strategy can be used to select the morphological filter for each block to optimize decoded image quality. Image quality can be enhanced by applying the appropriate morphological filter to each block. Experimental results demonstrate that the proposed technique enhances reconstructed ultrasound image quality compared to JPEG2000 at the same bit-rate in terms of both PSNR and the perceptual results.

A technique to reduce the overhead of vector quantization (VQ) is developed. This method exploits the high correlation property among neighboring blocks. By rearranging the codebook after the training process, the indices of neighboring vectors have fewer varieties, then, the gains of compression is attained. For the energy compactness advantage of the discrete cosine transform (DCT), the indices buffer formed after the encoding phase are fed to the DCT to reduce the bit rate further. Statistics illustrators are addressed to prove the efficient performance of the proposed method. Experiments are carried out on natural gray images. Simulation results show that our method decreases the bit rate more than 50% with the same reconstructed image quality compared to standard VQ coding.

Superresolution (SR) algorithms have recently become a hot research topic. The main purpose of image upscaling is to obtain high-resolution images from low-resolution ones, and these upscaled images should look like they had been taken with a camera having a resolution the same as the upscaled images, and at least present natural textures. In general, some SR algorithms preserve clear edges but blur the textures, while others preserve detailed textures but cause some obvious artifacts along edges. The proposed SR algorithm presents the detailed textures and, meanwhile, refines the strong edges and avoids causing obvious artifacts. The goal is achieved by using orthogonal fractal as the preliminary upscaling method in conjunction with the proper postprocessing where directional enhancement is adopted. In fact, the postprocessing part in the proposed SR algorithm can effectively reduce most jagged artifacts caused by SR algorithms. In the simulation results, it is shown that the proposed SR algorithm performs well in both objective and subjective measurements. Moreover, most detailed textures are properly enhanced and most jagged artifacts caused by SR algorithms can also be effectively reduced.

A good superresolution (SR) algorithm obtains high-resolution (HR) images from the corresponding low-resolution (LR) ones and, moreover, makes the former look like they had been acquired with a sensor having the expected resolution or at least as “natural” as possible. In general, fast SR algorithms usually result in more ill artifacts in the enlarged image, while the well-performed ones usually have great complexity and take much more computing time. For this purpose, four efficient SR algorithms based on regression models are proposed. In the proposed SR algorithms, the difference of a natural HR image and an HR image obtained by fast interpolation is taken as the lost detail and is supposed to be composed of several different oriented details. By the self-similarity of the input LR image and its corresponding HR image, a regression model is established by the input LR image to decide the proper respective weights of these oriented details which is then used to reconstruct the lost detail of the natural HR image. As shown in the experimental results, the proposed SR algorithms not only perform well in both objective criteria and visual quality but also take less computing time than some well-performing algorithms.

In this study, the authors present a new strategy to implement an illumination compensation-based contrast enhancement. Different from the traditional pixel-to-pixel transformation, the proposed method offers a level-to-level framework by generating the reference intensity level and the given target intensity level. Fundamentally, the traditional illumination compensation algorithms such as Histogram equalisation, log and gamma transformation are trade-off strategies and face the same dilemma. For example, the log function compresses the dynamic range of image with large variations in pixel values. The proposed method is an integration algorithm, the illumination compensation could be considered as the transformation from reference intensity to the chosen target intensity, and the contrast is enhanced by the given algebraic definition of the characteristic values of each pixel. The advantages are the adjusted intensity would be accomplished in the native phase according to the variant characteristic intensity values, clarifying the details of the darker areas as well as preserving the highlight areas, reducing HALO, colour constancy, removing colour cast and colour restoration. Shown as the experimental results, the proposed method can obtain better performance compared with other methods in subjective evaluation with contour plot and the objective evaluations by universal quality index, structural similarity index and average of standard deviations.

A fast vector quantization (VQ) encoder for video signals is developed. The video signal is encoded by VQ in a frame by frame manner. The VQ encoder can quickly search the codebook for a minimum distortion codeword by exploiting the spatial and temporal redundancy as well as using the dimension reduction technique. Finally a simplified labels replenishment technique is employed for bit rate reduction. >

Color demosaicking is critical to the image quality of single-sensor-based imaging devices. Caused by the sampling pattern of color filter array (CFA), the demosaicked images typically suffer from visual color artifacts in regions of high frequency and sharp edge structures, degrading the quality of camera output. We present a new high-quality demosaicking algorithm by taking advantage of deinterlacing and median-based filtering techniques. We treat the sampled green data of Bayer CFA as a form of diagonal interlaced green planes and make use of some key concepts about spatial deinterlacing to help the edge estimation in terms of both various directions and accuracy. In addition, a specific edge feature, sharp line edge of width 1 pixel, can also be handed well by the proposed method. The median-based filtering techniques are developed for suppressing most visual demosaicking artifacts, such as zipper effect, false color artifact, and interpolation artifact. Experimental results show that our algorithm is effective in suppressing visual artifacts, preserving the edges of image with sharpness and satisfying visual inspection, while keeping computational efficiency.

Most existing video compression standards use block-matching motion estimation to exploit temporal correlation between frames. To reduce the costly computation of exhaustive searches on all possible motion displacements, researchers have developed many fast algorithms. Some of these successfully proposed algorithms take advantage of the motion correlations between adjacent macroblocks. On the basis of the idea of second-order motion correlation between macroblocks in particular, we developed a new algorithm that applies a set of adaptive search patterns to benefit the statistical model of motions. In addition, adaptive early termination rules are used to prevent the waste of unnecessary computation. Simulation results show that the proposed algorithm outperforms most other existing algorithms in the areas of speed and visual quality.

A new method to remove blocking artefacts in low bit-rate block-based video coding, such as MPEG-4 and H.264, is presented. A low computational deblocking filter with five modes is proposed, including three frequency-related modes (smooth mode, intermediate mode, and complex mode for low-frequency, mid-frequency, and high-frequency regions, respectively), one special mode (steep mode for a large offset between two blocks) and a refined mode (corner mode for the corner of four blocks). A mode decision procedure is also needed to decide which mode is given by observing pixel behaviour around the block boundary. To take the masking effect of the human visual system (HVS) into consideration, the filter for smooth mode is designed to be much stronger than that for complex mode, because human eyes are more sensitive to smooth regions. Experimental results show that, in most cases, the proposed algorithm removes more blocking artefacts than MPEG-4 deblocking filters do, and improves both subjective and objective image quality. The proposed algorithm keeps the computation lower than MPEG-4 and is suitable for most block-based image and video coding systems.

The discrete cosine transform (DCT) is widely used in many practical image/video compression systems because of its compression performance and computational efficiency. We adopt the DCT and the modified set partitioning in hierachical trees (SPIHT) algorithm that was designed initially for encoding the dis- crete wavelet transform (DWT) coefficients to be suitable to encode DCT coefficients. The algorithm represents the DCT coefficients to concentrate signal energy and proposes a combination and dictator to eliminate the correlation in the same level subband for encoding DCT-based images. To further save bits, subbands with significant coefficients are classified into seven types. The coding complexity of the proposed algorithm for DCT coefficients is just close to JPEG but the performance is higher than JPEG2000. Experimental results indicate that the proposed technique improves the quality of the reconstructed image in terms of peak SNR (PSNR) over SPIHT and JPEG2000 at the same bit rate. © 2005 SPIE and