Krishna Ratakonda

POCS based adaptive image magnification

We tackle the problem of magnifying an image without incurring blurring, ringing or other artifacts common to classical schemes. The proposed iterative scheme starts with an initial magnified image generated by a process of selective interpolation. By placing suitable constraints on the final magnified image, which are convex in nature, we show that magnification can be posed as a problem of finding a solution which lies at the intersection of convex sets. By avoiding explicit high frequency enhancing assumptions in the iterative process, we avoid edge enhancement artifacts in the magnified image.

Hierarchical video summarization

We address the problem of key-frame summarization of vide in the absence of any a priori information about its content. This is a common problem that is encountered in home videos. We propose a hierarchical key-frame summarization algorithm where a coarse-to-fine key-frame summary is generated. A hierarchical key-frame summary facilitates multi-level browsing where the user can quickly discover the content of the video by accessing its coarsest but most compact summary and then view a desired segment of the video with increasingly more detail. At the finest level, the summary is generated on the basis of color features of video frames, using an extension of a recently proposed key-frame extraction algorithm. The finest level key-frames are recursively clustered using a novel pairwise K-means clustering approach with temporal consecutiveness constraint. We also address summarization of MPEG-2 compressed video without fully decoding the bitstream. We also propose efficient mechanisms that facilitate decoding the video when the hierarchical summary is utilized in browsing and playback of video segments starting at selected key-frames.

Robust video shot change detection

We present a novel improvement to existing schemes for abrupt shot change detection. Existing schemes declare a shot change whenever the frame to frame histogram difference (FFD) value is above a particular threshold. In such an approach, a high value for the threshold results in a small number of false alarms and a large number of missed detections while a low value for the threshold decreases the number of missed detections at the expense of increasing the false alarms. We attribute this situation to the fact that the FFD cannot be reliably used as the sole indicator for the presence of a shot change. In the proposed method a two-step shot detection strategy is used which selectively uses a likelihood ratio (computed directly from the frames and not from the histograms) to confirm the presence of a shot change. Such a two-step checking increases the probability of detection without increasing the probability of false alarm. The improvement proposed is simple and computationally cheap. Tests with a wide variety of video sequences prove the efficacy of the proposed approach.

Lossless image compression with multiscale segmentation

This paper is concerned with developing a lossless image compression method which employs an optimal amount of segmentation information to exploit spatial redundancies inherent in image data. Multiscale segmentation is obtained using a previously proposed transform which provides a tree-structured segmentation of the image into regions characterized by grayscale homogeneity. In the proposed algorithm we prune the tree to control the size and number of regions thus obtaining a rate-optimal balance between the overhead inherent in coding the segmented data and the coding gain that we derive from it. Another novelty of the proposed approach is that we use an image model comprising separate descriptions of pixels lying near the edges of a region and those lying in the interior. Results show that the proposed algorithm can provide performance comparable to the best available methods and 15-20% better compression when compared with the JPEG lossless compression standard for a wide range of images.

Image denoising using multiple compaction domains

We present a novel framework for denoising signals from their compact representation in multiple domains. Each domain captures, uniquely, certain signal characteristics better than others. We define confidence sets around data in each domain and find sparse estimates that lie in the intersection of these sets, using a POCS algorithm. Simulations demonstrate the superior nature of the reconstruction (both in terms of mean-square error and perceptual quality) in comparison to the adaptive Wiener filter.

MPEG-4 one-pass VBR rate control for digital storage

One-pass, variable bit-rate (VBR) rate control is ideally suited to the requirements of real-time video encoding for the purpose of digital storage. Previous MPEG one-pass VBR rate control algorithms have been based on appropriate selection of quantization scale parameters for controlling the bit rate and quality of the output bitstream. The major disadvantage of relying solely on quantization scales, for rate control, is the introduction of significant perceptual distortion when high quantization scales are used. We propose an MPEG-4, 1-pass, VBR rate control scheme that relies on the selective use of the MPEG-4 reduced resolution mode to supplement modulation of the quantization scale and provide an effective rate control strategy. Experimental results show that the proposed algorithm can encode high-complexity, standard definition (720 /spl times/ 480) video sequences at rates as low as 750 kbps without incurring significant perceptual artifacts.

Time-efficient learning theoretic algorithms for H.264 mode selection

The H.264 video coding standard derives much of its compression efficiency gain from the use of multiple different macroblock prediction modes for macroblock coding. In general, finding the prediction mode which gives optimal R-D performance for a given macroblock requires the encoder to completely encode the macroblock using all possible prediction modes. This results in a significant increase in encoder computational complexity. In this paper, we present a mode selection framework for H.264 which uses learning theoretic classification algorithms to discern between broad mode classes, based on the evaluation of a simple set of macroblock features. We show that the proposed mode selection framework significantly reduces encoder computational complexity, at the cost of only a small loss in compression performance.

Agility of Enterprise Operations across Distributed Organizations: A Model of Cross Enterprise Collaboration

We discuss the need for agility of business operations in a collaborative services ecosystem of partners and providers, and present a new system architecture for cross collaboration among multiple service enterprises. We demonstrate the importance and inevitability of such collaboration along with challenges in its proper realization through several real-life examples taken from different business domains. We then show that these challenges are rooted in two key factors: unpredictability and responsiveness; agility enables optimal response to unpredictable events. The key contribution of this manuscript is the presentation of a new model, centered on the modeling of work-asa-service (WaaS) and an intelligent hub for coordinating cross enterprise collaboration. This hub is constructed in a manner intended to directly identify and solve the two key fundamental challenges of cross enterprise collaboration. As such, we expect it to outperform other means of collaboration across service providers. We demonstrate the potential for such performance using field examples.

Quantitative Modeling of Communication Cost for Global Service Delivery

IT service providers are increasingly utilizing globally distributed resources to drive down costs, reduce risk through diversification and gain access to a larger talent pool. However, fostering effective collaboration among geographically distributed resources is a difficult challenge. In this paper, we present our initial attempt to quantify the increased overhead in leveraging distributed resources as one of the project costs. We associate this overhead cost measurement with metrics that measure communication quality, such as reduction in productivity and communication delay. These metrics can in turn be computed as functions of underlying project parameters. To achieve this goal, we first build a project communication model (PCM) to categorize different types of collaborative communication. We then represent communication efficiency and changes in resource availability in terms of information theoretic concepts such as reduced channel capacity, information encoding efficiency and channel availability. This analysis is used to help determine the cost associated with team formation and task distribution during the project planning phase.

Multiple description coding of predictively encoded sequences

We are concerned with multiple description coding of predictively encoded sequences at low rates/redundancies. The key problem to be solved is that of avoiding predictive mismatch. We propose an algorithm based on multiple description correlating transforms (MDCT), which avoids the problem of predictive mismatch while not increasing the number of transmitted coefficients. Under a general first-order, Gauss-Markov source model, carefully formulated to be applicable to the practically important case of video coding, we obtain closed-form expressions for the operational rate-distortion characteristics of the proposed algorithm and use these to optimize the algorithm encoding parameters. In the case of the residual-of-residuals technique, which is also based on MDCT, we show that incorporation of any redundancy between the transmitted descriptions is inherently suboptimal. Results show that the proposed algorithm significantly outperforms the residual-of-residuals technique at low rates.