Efficient Belief Propagation in Depth Finding (original) (raw)
Related papers
Accelerated Belief Propagation for Hardware Implementation
Disparity map plays an important role in 3DTV and FTV systems. Despite recent advances, state-of-the-art algorithms fail to generate a precise disparity map rapidly enough for VLSI real time processing. Belief Propagation (BP) is a popular global optimization algorithm and has several advantages for hardware implementation. However, it requires high bandwidth, memory and computational costs. Tile-based BP is an efficient improved BP for hardware implementation. Boundary messages loaded from different directions have been performed with different iterations. As a result, energy cost of different pixels within a tile will converge with different speed, yielding the biasing problem. In this paper, firstly, a Balanceconvergence method to remove biasing problem in tile-based BP is presented. Additionally, a novel reduced message-update method to remove redundant computational costs during message-update based on tile-based BP is introduced. Compared with the original BP and the leading hardware-oriented fast method, the proposed method can reduce additions by 30 times and 2 times respectively.
2009
3D technologies are becoming the more and more relevant in recent years. Visual communications, as well as image and video analysis, benefit in great manner from spatial information such as depth for various applications. Highly accurate visual depth estimation often involves complex optimization algorithms in order to fit proper estimation models to data. From a stereo/multiview matching perspective, local and global algorithms exist. Commonly, the latter are more complex and accurate, as data models are used to take the global structure into account. Belief Propagation has proven to be a good global algorithmic framework for depth estimation. By means of an iterative procedure, it is able to regularize, according to set of local smoothness and geometry constrains, an initial estimation of depth by a local approach such as simple block matching. However, information transfer from iteration to iteration by means of message passing can be excessively demanding in terms of memory bandwidth and usage. In this paper, a new Belief Propagation based algorithm with multiview matching with depth/color segmentation is proposed together with a strategy for message passing compression. Experimental results show the algorithm to be competitive with best performing ones in the state of the art, while reducing by a factor 10 the memory usage, with marginal loss in performance, of a typical Belief Propagation strategy.
Photogrammetric Engineering & Remote Sensing, 2009
Pixel-swapping algorithm is a simple and efficient technique for sub-pixel mapping (Atkinson, 2001 and 2005). It was initially applied in shoreline and rural land-cover mapping but has been expanded to other land-cover mapping. However, due to its random initializing process, this algorithm must swap a large number of sub-pixels, and therefore it is computation intensive. This computing power consumption intensifies when the scale factor is large. A new, modified pixel-swapping algorithm (MPS) is presented in this paper to reduce the computation time, as well as to improve sub-pixel mapping accuracy. The MPS algorithm replaces the original random initializing process with a process based on a sub-pixel/pixel spatial attraction model. The new algorithm was used to allocate multiple land-covers at the subpixel level. The results showed that the MPS algorithm outperformed the original algorithm both in sub-pixel mapping accuracy and computational time. The improvement is especially significant in the case of large scale factors. Furthermore, the MPS is less sensitive to the size of neighboring sub-pixels and can still result in increased accuracy even if the size of neighbors is small. The MPS was also much less time consuming, as it reduced both the iterations and total amount of swapping needed.
Efficient Stereo Algorithm using Multiscale Belief Propagation on Segmented Images
British Machine Vision Conference, 2008
A variety of approaches using BP and image segmentation have been proposed for the stereo correspondence problem. In this paper, we introduce a novel approach, based on a combination of segmentation and BP. Our method inherits the idea of Multiscale BP, however at each level of the hierarchy, each graph node corresponds to an image segment, which we call superpixel, instead of a fixed rectangular block of pixels. The resulting depth map at the coarser level is used to initialize the depths at the finer level. At the lowest level, we perform loopy BP on the four-connected pixel subgrid within each superpixel. The proposed method is applied to stereo images in the standard Middlebury dataset, and to real outdoor stereo images and car sequences. Experimental results show quite acceptable accuracy of depth inference, with running time fast enough for practical use.
2007
The paper deals with the design and implementation of a stereo algorithm. Disparity map is formulated as a Markov Random Field with a new smoothness constraint depending not only on image derivatives, but also on segmentation results and gradient directions. With these constraints we force disparity continuity inside each segmented object, while its contours are well preserved. Moreover we have designed a modified version of Belief Propagation which gives the solution to the stereo matching problem: the optimization has remarkable improvements and especially with respect to message propagation, which is actually driven by segmentation and boundary knowledge. Preliminary results are presented both on synthetic and benchmark images to demonstrate the effectiveness of our method.
Real time disparity map estimation on the cell processor
Stereo vision attempts to regain depth information from a pair of 2D images. This information can be used in a wide range of areas including robotics, augmented reality, 3D TV and movie post production. This paper will describe the development of a stereo algorithm on the cell processor. The implementation will be evaluated for both quality and speed, demonstrating the effectiveness of the cell processor as an image processing platform. The restrictions imposed by the platform and issues which are faced when developing computer vision applications on the cell will be discussed and the methods taken to implement the algorithm will be described.
Towards the Optimal Hardware Architecture for Computer Vision
Machine Vision - Applications and Systems, 2012
After image acquisition, some preprocessing steps are often required. These are intended to provide reliable input data for subsequent computing stages. Some typical operations are 248 Machine Vision-Applications and Systems www.intechopen.com Towards the Optimal Hardware Architecture for Computer Vision 3 noise reduction, color balancing, geometrical transformation, etc. Most of these operations are based on point or near-neighborhood operations. Point operations are performed at pixel-level in such a way that the output only depends on the value of any individual pixels from one or several input images. With this type of operation it is possible to modify the pixel intensity to enhance parts of the image, by increasing contrast or brightness. Equally, simple pixel-to-pixel arithmetic and Boolean operations also enable the construction of operators as alpha blending, for image combination or color space conversion. Neighborhood operations take also into account the value of adjacent pixels. This operation type is the basis of filtering, binary morphology or geometric transformation. They are characterized by simple operations, typically combining weighted sums, Boolean and thresholding processing steps.
Low Memory Cost Block-Based Belief Propagation for Stereo Correspondence
Multimedia and Expo, 2007 IEEE International Conference on, 2007
The typical belief propagation has good accuracy for stereo correspondence but suffers from large run-time memory cost. In this paper, we propose a block-based belief propagation algorithm for stereo correspondence that partitions an image into regular blocks for optimization. With independently partitioned blocks, the required memory size could be reduced significantly by 99% with slightly degraded performance with a 32x32 block size when compared to original one. Besides, such blocks are also suitable for parallel hardware implementation. Experimental results using Middlebury stereo test bed demonstrate the performance of the proposed method.
Efficient Belief Propagation for Early Vision
such as stereo and image restoration. Inference algorithms based on graph cuts and belief propagation have been found to yield accurate results, but despite recent advances are often too slow for practical use. In this paper we present some algorithmic techniques that substantially improve the running time of the loopy belief propagation approach. One of the techniques reduces the complexity of the inference algorithm to be linear rather than quadratic in the number of possible labels for each pixel, which is important for problems such as image restoration that have a large label set.
Real-time Global Stereo Matching Using Hierarchical Belief Propagation
Procedings of the British Machine Vision Conference 2006
In this paper, we present a belief propagation based global algorithm that generates high quality results while maintaining real-time performance. To our knowledge, it is the first BP based global method that runs at real-time speed. Our efficiency performance gains mainly from the parallelism of graphics hardware,which leads to a 45 times speedup compared to the CPU implementation. To qualify the accurancy of our approach, the experimental results are evaluated on the Middlebury data sets, showing that our approach is among the best (ranked first in the new evaluation system) for all real-time approaches. In addition, since the running time of general BP is linear to the number of iterations, adopting a large number of iterations is not feasible for practical applications. Hence a novel approach is proposed to adaptively update pixel cost. Unlike general BP methods, the running time of our proposed algorithm dramatically converges.