REal-time local stereo matching using guided image filtering

Abstract

Adaptive support weight algorithms represent the state-of the-art in local stereo matching. Their limitation is a high computational demand, which makes them unattractive for many (real-time) applications. To our knowledge, the algo rithm proposed in this paper is the first local method which is both fast (real-time) and produces results comparable to global algorithms. A key insight is that the aggregation step of adaptive support weight algorithms is equivalent to smoothing the stereo cost volume with an edge-preserving filter. From this perspective, the original adaptive support weight algo rithm [1] applies bilateral filtering on cost volume slices, and the reason for its poor computational behavior is that bilat eral filtering is a relatively slow process. We suggest to use the recently proposed guided filter [2] to overcome this limi tation. Analogously to the bilateral filter, this filter has edge preserving properties, but can be implemented in a very fast way, which makes our stereo algorithm independent of the size of the match window. The GPU implementation of our stereo algorithm can process stereo images with a resolution of 640 x 480 pixels and a disparity range of 26 pixels at 25 fps. According to the Middlebury on-line ranking, our algo rithm achieves rank 14 out of over 100 submissions and is not only the best performing local stereo matching method, but also the best performing real-time method.

Runtime Performance and Power Optimization of Parallel Disparity Estimation on Many-Core Platforms

“…Recently, the use of a Guided Image Filter (GIF) [He et al 2013] in local ADSW algorithms has been proposed to reduce the complexity of cost aggregation, leading to a high-quality, fast and simple local DE algorithm [Hosni et al 2011]. Due to the reduced complexity of this type of filter, the algorithm can operate at real-time framerates for HD images when implemented in a parallel structure [Ttofis et al 2016].…”

“…These variables are derived from the same literature as the algorithm and are set to the constant values: {α, T c , T g } = {0.9, 0.028, 0.008} [Hosni et al 2011]. The parameters were found empirically by those authors who do not provide further information on how they were derived.…”

“…Cost Volume Construction (CVC) is the comparison of pixels between the two images over a range of disparities d. A cost value is assigned to each pixel p in the left image based on the dissimilarity between it and a pixel in the right image. A truncated absolute difference of colors (M (p, d), equation (1)) and gradients (G(p, d), equation (2)) are the cost contributions, at pixel p and disparity d [Hosni et al 2011]. I is the value of each pixel color channel in equation (1) and the luminance in equation (2).…”

Cost volume-based interactive depth editing in stereo post-processing

“…Our OpenCL implementation uses a tile-based sliding-window variant which works in O(n) on the GPU [6].…”

Robust stereo matching with trinary cross color census and triple image-based refinements

“…With multiple cameras [10,11], the stereo vision technologies [12][13][14] to extract the depth information become a lowprice and high-resolution approach. With horizontally placed cameras, the distance estimation of each pixel, called stereo matching, searches the best correspondence of the same scene point in two different viewing images [15,16]. The horizontal displacement of the paired pixels in two viewing images is called the disparity.…”