Stereo matching algorithm based on illumination normal similarity and adaptive support weight

Abstract

Abstract. For the purpose of representing the feature of the gray image, illumination normal of pixels in a two-dimensional gray image plane is proposed, which can reflect the high-frequency information of the gray image. In order to get an accurate dense disparity map based on the adaptive support weight (ASW) approach in RGB vector space, a matching algorithm is proposed that combines the illumination normal similarity, gradient similarity, color similarity, and Euclidean distance similarity to compute the corresponding support weights and dissimilarity measurements. After testing by the Middlebury stereo benchmark, the result of the proposed algorithm shows more accurate disparity than many state-of-the-art stereo matching algorithms.

Augmented Reality Implementations in Stomatology

“…This enables locations of the teeth to be evaluated during treatment from the spatial coordinates that have been obtained. This transform allows us to detect changes in the dental arch with a stereo matching algorithm [38,39]. Corresponding pairs of images obtained before and after the dental treatment enable us to use 2D registration to evaluate the results of the operations [40].…”

Accurate Dense Stereo Matching Based on Image Segmentation Using an Adaptive Multi-Cost Approach

“…The non-parametric transform-based similarity measure function is more robust to radiometric distortion and noise than the intensity based. For cost aggregation aspect, the adaptive window [18][19][20] and adaptive weight [17,21,22] are two principal methods. Adaptive window methods try to assign an appropriate size and shape support region for the given pixel to aggregate the raw costs.…”

“…Two of them are traditional similarity functions, which are absolute difference similarity functions that take into account information from RGB (Red, Green, Blue) channels, and the similarity function based on the principal image gradients. The other two similarity functions are improved census transform [7] and illumination normal vector [22]. An efficient adaptive method of aggregating initial matching cost for each pixel is then applied, which relies on a linearly expanded cross skeleton support window.…”

“…INV reflects the high-frequency information of the image, which generally exists at the boundaries of objects and fine texture area. Consequently, the high-frequency information reflects some small-scale details of the image, which is very useful for stereo correspondence [22]. Denote a pixel of the image as a point in 3D space P[x, y, f (x, y)], where x and y are the horizontal and vertical coordinates, and f (x,y) is the intensity value of position (x,y).…”