Ranjith Unnikrishnan

Toward Objective Evaluation of Image Segmentation Algorithms

Unsupervised image segmentation is an important component in many image understanding algorithms and practical vision systems. However, evaluation of segmentation algorithms thus far has been largely subjective, leaving a system designer to judge the effectiveness of a technique based only on intuition and results in the form of a few example segmented images. This is largely due to image segmentation being an ill-defined problem-there is no unique ground-truth segmentation of an image against which the output of an algorithm may be compared. This paper demonstrates how a recently proposed measure of similarity, the normalized probabilistic rand (NPR) index, can be used to perform a quantitative comparison between image segmentation algorithms using a hand-labeled set of ground-truth segmentations. We show that the measure allows principled comparisons between segmentations created by different algorithms, as well as segmentations on different images. We outline a procedure for algorithm evaluation through an example evaluation of some familiar algorithms - the mean-shift-based algorithm, an efficient graph-based segmentation algorithm, a hybrid algorithm that combines the strengths of both methods, and expectation maximization. Results are presented on the 300 images in the publicly available Berkeley segmentation data set

A Measure for Objective Evaluation of Image Segmentation Algorithms

Despite significant advances in image segmentation techniques, evaluation of these techniques thus far has been largely subjective. Typically, the effectiveness of a new algorithm is demonstrated only by the presentation of a few segmented images and is otherwise left to subjective evaluation by the reader. Little effort has been spent on the design of perceptually correct measures to compare an automatic segmentation of an image to a set of hand-segmented examples of the same image. This paper demonstrates how a modification of the Rand index, the Normalized Probabilistic Rand (NPR) index, meets the requirements of largescale performance evaluation of image segmentation. We show that the measure has a clear probabilistic interpretation as the maximum likelihood estimator of an underlying Gibbs model, can be correctly normalized to account for the inherent similarity in a set of ground truth images, and can be computed efficiently for large datasets. Results are presented on images from the publicly available Berkeley Segmentation dataset.

Measures of Similarity

Quantitative evaluation and comparison of image segmentation algorithms is now feasible owing to the recent availability of collections of hand-labeled images. However, little attention has been paid to the design of measures to compare one segmentation result to one or more manual segmentations of the same image. Existing measures in statistics and computer vision literature suffer either from intolerance to labeling refinement, making them unsuitable for image segmentation, or from the existence of degenerate cases, making the process of training algorithms using the measures to be prone to failure. This paper surveys previous work on measures of similarity and illustrates scenarios where they are applicable for performance evaluation in computer vision. For the image segmentation problem, we propose a measure that addresses the above concerns and has desirable properties such as accommodation of labeling errors at segment boundaries, region sensitive refinement, and compensation for differences in segment ambiguity between images

Multi-scale interest regions from unorganized point clouds

Several computer vision algorithms rely on detecting a compact but representative set of interest regions and their associated descriptors from input data. When the input is in the form of an unorganized 3D point cloud, current practice is to compute shape descriptors either exhaustively or at randomly chosen locations using one or more preset neighborhood sizes. Such a strategy ignores the relative variation in the spatial extent of geometric structures and also risks introducing redundancy in the representation. This paper pursues multi-scale operators on point clouds that allow detection of interest regions whose locations as well as spatial extent are completely data-driven. The approach distinguishes itself from related work by operating directly in the input 3D space without assuming an available polygon mesh or resorting to an intermediate global 2D parameterization. Results are shown to demonstrate the utility and robustness of the proposed method.

Vegetation Detection for Driving in Complex Environments

A key challenge for autonomous navigation in cluttered outdoor environments is the reliable discrimination between obstacles that must be avoided at all costs, and lesser obstacles which the robot can drive over if necessary. Chlorophyll-rich vegetation in particular is often not an obstacle to a capable off-road vehicle, and it has long been recognized in the satellite imaging community that a simple comparison of the red and near-infrared (NIR) reflectance of a material provides a reliable technique for measuring chlorophyll content in natural scenes. This paper evaluates the effectiveness of using this chlorophyll-detection technique to improve autonomous navigation in natural, off-road environments. We demonstrate through extensive experiments that this feature has properties complementary to the color and shape descriptors traditionally used for point cloud analysis, and show significant improvement in classification performance for tasks relevant to outdoor navigation. Results are shown from field testing onboard a robot operating in off-road terrain.

Scale selection for classification of point-sampled 3D surfaces

Three-dimensional ladar data are commonly used to perform scene understanding for outdoor mobile robots, specifically in natural terrain. One effective method is to classify points using features based on local point cloud distribution into surfaces, linear structures or clutter volumes. But the local features are computed using 3D points within a support-volume. Local and global point density variations and the presence of multiple manifolds make the problem of selecting the size of this support volume, or scale, challenging. In this paper, we adopt an approach inspired by recent developments in computational geometry (Mitra et al., 2005) and investigate the problem of automatic data-driven scale selection to improve point cloud classification. The approach is validated with results using data from different sensors in various environments classified into different terrain types (vegetation, solid surface and linear structure).

Robust extraction of multiple structures from non-uniformly sampled data

The extraction of multiple coherent structures from point clouds is crucial to the problem of scene modeling. While many statistical methods exist for robust estimation from noisy data, they are inadequate for addressing issues of scale, semi-structured clutter, and large point density variation together with the computational restriction of autonomous navigation. This paper extends an approach of nonparametric projection-pursuit based regression to compensate for the non-uniform and directional nature of data sampled in outdoor environments. The proposed algorithm is employed for extraction of planar structures and clutter grouping. Results are shown for scene abstraction of 3D range data in large urban scenes.

Extracting Scale and Illuminant Invariant Regions through Color

Despite the fact that color is a powerful cue in object recognition, the extraction of scale-invariant interest regions from color images frequently begins with a conversion of the image to grayscale. The isolation of interest points is then completely determined by luminance, and the use of color is deferred to the stage of descriptor formation. This seemingly innocuous conversion to grayscale is known to suppress saliency and can lead to representative regions being undetected by procedures based only on luminance. Furthermore, grayscaled images of the same scene under even slightly different illuminants can appear sufficiently different as to affect the repeatability of detections across images. We propose a method that combines information from the color channels to drive the detection of scale-invariant keypoints. By factoring out the local effect of the illuminant using an expressive linear model, we demonstrate robustness to a change in the illuminant without having to estimate its properties from the image. Results are shown on challenging images from two commonly used color constancy datasets.

A constrained optimization approach to globally consistent mapping

Mobile robot localization from large-scale appearance mosaics has been showing increasing promise as a low-cost, high-performance and infrastructure free solution to vehicle-guidance in man-made environments. The generation of the globally consistent high-resolution mosaics crucial to this procedure suffers from the same problem of loop-closure in cyclic environments that is commonly encountered in all map-building procedures. This paper presents a batch solution to the problem of reliably generating globally consistent mosaics at low computational cost, that simultaneously exploits the topological constraints among the observations and minimizes the total residual in observed features. An extension to a general scalable framework that facilitates an incremental online mapping strategy is also presented, along with results using simulated data and from real indoor environments.

Scale Selection for the Analysis of Point-Sampled Curves

An important task in the analysis and reconstruction of curvilinear structures from unorganized 3-D point samples is the estimation of tangent information at each data point. Its main challenges are in (1) the selection of an appropriate scale of analysis to accommodate noise, density variation and sparsity in the data, and in (2) the formulation of a model and associated objective function that correctly expresses their effects. We pose this problem as one of estimating the neighborhood size for which the principal eigenvector of the data scatter matrix is best aligned with the true tangent of the curve, in a probabilistic sense. We analyze the perturbation on the direction of the eigenvector due to finite samples and noise using the expected statistics of the scatter matrix estimators, and employ a simple iterative procedure to choose the optimal neighborhood size. Experiments on synthetic and real data validate the behavior predicted by the model, and show competitive performance and improved stability over leading polynomial-fitting alternatives that require a preset scale.