Search

This paper presents an efficient and effective interactive segmentation scheme for extracting the region of a foreground object in an image. Our goal is to design an interactive segmentation algorithm that unifies the bounding-box-based, seed-based, and query-based interaction mechanisms for pursuing (i) high efficiency in simple interaction mechanism, (ii) few interaction rounds, and (iii) short response time. The proposed algorithm starts with a user-provided bounding box and obtains candidate background superpixels for inferring the foreground object. Our algorithm tolerates imprecise bounding boxes and provides two kinds of interactions for acquiring correct labels from the user. The user can either input the seed/scribble annotations or label the algorithm-queried regions. Our algorithm selects the most uncertain region as a query, and this query-based interaction mechanism reduces the burden of the user on deciding suitable annotation locations. The average response time per-interaction of our algorithm is merely 0.014 s. Our experiments demonstrate that the algorithm achieves an efficient unified scheme for interactive image segmentation.

This paper presents an effective algorithm, interactive 1-bit feedback segmentation using transductive inference (FSTI), that interactively reasons out image segmentation. In each round of interaction, FSTI queries the user one superpixel for acquiring 1-bit user feedback to define the label of that superpixel. The labeled superpixels collected so far are used to refine the segmentation and generate the next query. The key insight is treating the interactive segmentation as a transductive inference problem, and then suppressing the unnecessary queries via an intrinsic-graph-structure derived from transductive inference. The experiments conducted on five publicly available datasets show that selecting query superpixels concerning the intrinsic-graph-structure is helpful to improve the segmentation accuracy. In addition, an efficient boundary refinement is presented to improve segmentation quality by revising the misaligned boundaries of superpixels. The proposed FSTI algorithm provides a superior solution to the interactive image segmentation problem is evident.

This paper proposes a secure reversible visible watermarking approach. The proposed pixel mapping function superposes a binary watermark image on a host image to create an intermediate visible watermarked image. Meanwhile, an almost inverse function generates the recovery data for restoring the original pixels. To prevent unauthorized users from approximating the original pixels in the watermarked region, this method adds an integer sequence in the intermediate watermarked image. The sequence is composed of integers generated by two random variables having normal distributions with zero means and distinct variances. The variances facilitate a trade-off between the watermark transparency and the noise generated by unauthorized users. The proposed method also uses Lagrange multipliers to find the optimized variances for the trade-off. Finally, this method uses reversible data embedding to embed the recovery data and hash value for reversibility and authentication, respectively. Experimental results show the watermark visibility for test images along with the watermark transparency for different variances. Using the optimized variances, the watermarked image is at the balance between the watermark transparency and the unauthorized-user-generating noise.

Image matching has been a central problem in computer vision and image processing for decades. Most of the previous approaches to image matching can be categorized into the intensity-based and edge-based comparison. Hausdorff distance has been widely used for comparing point sets or edge maps since it does not require point correspondences. In this paper, we propose a new image similarity measure combining the Hausdorff distance with a normalized gradient consistency score for image matching. The normalized gradient consistency score is designed to compare the normalized image gradient fields between two images to alleviate the illumination variation problem in image matching. By combining the edge-based and intensity-based information for image matching, we are able to achieve robust image matching under different lighting conditions. We show the superior robustness property of the proposed image matching technique through experiments on face recognition under different lighting conditions.

Bit rate control has been proved to be an important technique used to transmit the variable rate bit stream over a fixed rate channel in MPEG coding. In this paper, a hybrid algorithm is proposed to achieve the rate control and obtain a good picture quality. In the picture layer, the algorithm estimates numbers of bits required for different type of frames. In the macroblock layer, it utilizes the predictive and adaptive perceptual techniques in the regulation of the quantization scale to allocate the target bits. In the upper and left boundary of the picture, the macroblocks are quantized with the perceptual basis to adjust the quantization scale and in the other part of the picture a predictive method is used to update the quantization scale. By using the proposed scheme, the distortion of coding is uniformly distributed in the picture layer as well as in the macroblock layer and better bit rate control is achieved.

A new adaptive post-processing algorithm for the MPEG decoded video sequences is proposed. We use a motion compensated averaging filter to reduce the noises in the temporal domain and an adaptive spatial filter to remove noise in the spatial domain and preserve the edge of different orientations in the image. A MPEG decoded video sequence called table tennis is processed by our proposed filter. The post-processed video sequence shows that its image quality is improved, especially of the moving objects.

The compression of Chinese characters is very important for Chinese office automation and desktop publishing. In this paper, various methods are used to compress 13,051 Chinese characters losslessly by coding their skeleton points. The skeleton points are composed of isolated points and curves. To trace every point in a curve once and consecutively is a NP problem. Therefore, we use contour path following and revisiting algorithms to trace skeleton curves with chain codes. The remaining isolated skeleton points are encoded by an Elias code. Our coding method has a better compression rate than that of conventional skeleton coding.

Chain codes [Freeman 1974; Charavarty 1982; Kim, Lee and Kim 1988; Morrin 1976; Pavlidis 1978] and y-axis representation [Merrill 1973] are both commonly used representation of a region in a binary image. The y-axis representation is very efficient for some geometric applications [Chang and Leu 1990] and morphological operations [Ji, Piper and Tang 1989]. However, it is not efficient to compute the properties on the contour. In this paper, a new and simple algorithm other than Kim's [Kim, Lee and Kim 1988] to convert the y-axis representation to a sequence of chain codes is developed.

The order statistic (OS) filter of M-level signals has three stages: thresholding, binary filtering, and reconstruction. For binary filtering, the authors use a pipelined sorting network instead of positive Boolean functions, which are very complex for a generalized OS filter. They also develop a fast reconstructor with (2M-2-log/sub 2/M) half adders. Both the computation time and latency time are just four times that of one gate delay. The design is also suitable for VLSI cellular array implementation. >

The authors present a systolic algorithm and its variations for the k-nearest neighbors problem (kNNP). Multiple-shot queries with different ranges (k values) can be served in a pipelined fashion. A partitioning scheme is developed to handle large size problems. Performance of the algorithm is analyzed. Formulas for the optimal array size in terms of computation time and area-time-time product (ATT) are derived. The algorithm can solve a multiple-shot kNNP in N+2 square root N*K systolic steps using square root N*K processing elements, where N is the problem size (i.e. the number of points), and K is the sum of all k-values. >

A bit-level systolic array for median filters based on the majority of the bit strings of input samples is presented. The majority function can be implemented by a 1-b odd/even transposition network, which is regular and can be designed easily. If each input sample is represented by r bits, the systolic array contains r processor elements and requires time complexity O(N+n+r-1) for a stream of N samples with window size n in a one-dimensional median filter. >

Recently, R.N. Bracewell (1983) introduced the discrete Hartley transform (DHT) as an alternative to the discrete Fourier transform (DFT). Two linear systolic array models for the (DHT) are derived. One model requires O(2N-1) in the computational phase and O(N) in the preloading phase. The other model requires O(2N-1) in the computational phase and O(N) in the output phase. A square systolic array for two-dimensional DHT is also constructed by combining the individual advantages of each model. The CORDIC algorithm is proposed as an alternative to conventional multipliers. To speed up the systolic array, two-level pipelining with CORDIC is also possible. >

A linear systolic array for the discrete cosine transform, discrete sine transform, and their inverses is developed. It generates the transform kernel values recursively. Compared to the scheme with the transform kernel values prestored in memory either inside or outside each processing element, the clock period is shortened by a memory access time. In addition, the array pays no cost for prestorage. The systolic array has the advantages of pipelinability, regularity, locality, and scalability, making it quite suitable for VLSI signal processing. >

A new algorithm for image sequence coding with a very low bit rate and high picture quality is proposed. The proposed cod- ing system is composed of interframe and intraframe coding. In the interframe coding, an edge-based motion estimation algorithm is used to find the motion vector of a moving block instead of using the conventional block-matching algorithm. The edge-based method re- duces the entropy of the estimated motion vectors within a frame. Since the motion vectors are usually encoded by entropy coding, bit rate is also reduced. In the intraframe coding, a region-based algo- rithm is used to trace the contour of an object and estimate its pa- rameters from a predetermined database of various geometric shapes. It can reduce the blocking effect. By testing several image sequences such as ''Miss America'' and ''Claire'' with 10 frames/s in quarter common intermediate format (QCIF) the proposed coding system has about 1 to 2 dBs higher peak SNRs than the H.263 system. © 1998 SPIE and IS&T. (S1017-9909(98)01401-9)

Face image matching is an essential step for face recognition and face verification. It is difficult to achieve robust face matching under various image acquisition conditions. In this paper, a novel face image matching algorithm robust against illumination variations is proposed. The proposed image matching algorithm is motivated by the characteristics of high image gradient along the face contours. We define a new consistency measure as the inner product between two normalized gradient vectors at the corresponding locations in two images. The normalized gradient is obtained by dividing the computed gradient vector by the corresponding locally maximal gradient magnitude. Then we compute the average consistency measures for all pairs of the corresponding face contour pixels to be the robust matching measure between two face images. To alleviate the problem due to shadow and intensity saturation, we introduce an intensity weighting function for each individual consistency measure to form a weighted average of the consistency measure. This robust consistency measure is further extended to integrate multiple face images of the same person captured under different illumination conditions, thus making our robust face matching algorithm. Experimental results of applying the proposed face image matching algorithm on some well-known face datasets are given in comparison with some existing face recognition methods. The results show that the proposed algorithm consistently outperforms other methods and achieves higher than 93% recognition rate with three reference images for different datasets under different lighting conditions.