Using Models of Objects with Deformable Parts for Joint Categorization and Segmentation of Objects

We propose to bridge the gap between Random Field (RF) formulations for joint categorization and segmentation (JCaS), which model local interactions among pixels and superpixels, and Bag of Features categorization algorithms, which use global descriptors. For this purpose, we introduce new higher order potentials that encode the classification cost of a histogram extracted from all the objects in an image that belong to a particular category, where the cost is given as the output of a classifier when applied to the histogram. The potentials efficiently encode the classification costs of several histograms resulting from the different possible segmentations of an image. They can be integrated with existing potentials, hence providing a natural unification of global and local interactions. The potentials' parameters can be treated as parameters of the RF and hence be jointly learnt along with the other parameters of the RF. Experiments show that our framework can be used to improve the performance of existing JCaS algorithms.

The final goal of this work is to automatically extract the contour of an object represented in an image after manually defining an initial contour for it. This rough initial contour will then evolve until it equals the border of the desired object. The contour is modelled by a physical formulation, using the Finite Element Method, and its evolution to the desired final contour of the object to segment is governed by several forces: internal forces, defined by intrinsic physical characteristics selected for the model; and external forces, defined in function of image features that best represent the desired object. To physically model the considered contour we adopt Sclaroff’s isoparametric finite element, and to obtain the evolution of the model towards the object border we use Nastar’s methodology that consists in solving the dynamic equilibrium equation.

Automatic rib segmentation and labeling in computed tomography scans using a general framework for detection, recognition and segmentation of objects in volumetric data

We present a novel form of interactive object segmentation called Click Carving which enables accurate segmentation of objects in images and videos with only a few point clicks. Whereas conventional interactive pipelines take the user’s initialization as a starting point, we show the value in the system taking lead even in initialization. In particular, for a given image or a video frame, the system precomputes a ranked list of thousands of possible segmentation hypotheses (also referred to as object region proposals) using appearance and motion cues. Then, the user looks at the top ranked proposals, and clicks on the object boundary to carve away erroneous ones. This process iterates (typically 2–3 times), and each time the system revises the top ranked proposal set, until the user is satisfied with a resulting segmentation mask. In the case of images, this mask is considered as the final object segmentation. However in the case of videos, the object region proposals rely on motion as well, and the resulting segmentation mask in the first frame is further propagated across the video to obtain a complete spatio-temporal object tube. On six challenging image and video datasets, we provide extensive comparisons with both existing work and simpler alternative methods. In all, the proposed Click Carving approach strikes an excellent of accuracy and human effort. It outperforms all similarly fast methods, and is competitive or better than those requiring 2–12 times the effort.

The Active Contour Model (ACM) was introduced in 1987 and allows the image segmentation of objects in digital images through edge detection, obtained from image information. This ACM consists in outlining an initial curve around or inside an object of interest. This curve is deformed, according to some forces which act on it, moving it to the edges of the object. This process is performed through successive minimizing iterations of a given energy associated to the curve. Recently, various applications have proposed changes in traditional ACMs, generating new methods, applications, benefits and performance improvements in certain applications. Due to that, these methods have also been expanded to three dimensions (3D), being called Active Contour Methods 3D (ACM 3D), as is the case of the proposed method, called Adaptive Balloon ACM 3D. This method is applied in the segmentation of 3D models and compared to the 3D Region Growing method (RG 3D). The results showed that the RG is superior to the 3D proposed method when applied in closed objects, but inferior in hollow objects, in which the proposed method is stably displayed in all applications. Therefore, one can conclude that the proposed method can be applied to the segmentation of 3D objects in a promising way.

Statistical neural networks executing soft-decision algorithms have been shown to be very effective in many classification problems. A neural network architecture is developed here that can perform unsupervised joint segmentation and labeling of objects in images. We propose the semi-parametric hierarchical mixture density (HMD) model as a tool for capturing the diversity of real world images and pose the object recognition problem as a maximum likelihood (ML) estimation of the HMD parameters. We apply the expectation-maximization (EM) algorithm for this purpose and utilize ideas and techniques from statistical physics to cast the problem as the minimization of a free energy function. We then proceed to regularize the solution thus obtained by adding smoothing terms to the objective function. The resulting recursive scheme for estimating the posterior probabilities of an object's presence in an image corresponds to an unsupervised feedback neural network architecture. We present here the results of experiments involving recognition of traffic signs in natural scenes using this technique.

In this paper, a generative model combined with stochastic framework is proposed and applied to the simultaneous correspondence estimation and object segmentation. The correspondence and segment fields are explicitly modelled as Markov random fields, and estimated in the maximum a posteriori framework. Some stochastic models are defined as the potential functions to reflect the interaction of the fields. The potential functions of the fields are stochastically diffused with the probability distributions of the neighboring fields, and the probability spaces of the fields are updated from the diffused potential spaces. The stochastic diffusion proposed as an energy minimization process is a kind of generative model which updates and regenerates the probability spaces of the correspondence and segment fields. Some experiments are performed on the simultaneous correspondence estimation and object segmentation. The results show stable and good performances in estimating the correspondence fields and extracting the objects in the scene.

Detection of particular meaningful objects in images is of great importance in many fields, including image retrieval or image quality analysis. In this contribution, eleven frequently observed objects (areas) in natural images are learned and detected. The presented algorithm is based on region merging and Bayesian decision theory. The main goal is not perfect recognition, since for that purpose it is necessary to use higher-level knowledge about the image content. Merging of segments proceeds only up to a reliable point, not to merge different categories. Unique merging criteria combine the values of probabilities attached to the segments for all the most likely categories, color and texture features and information about edges. Results are demonstrated on a few images and discussed.

Terahertz passive and active imaging technology make it possible to detect concealed objects under clothing at a standoff distance. Compared with passive THz imaging scheme, several factors in active THz imaging make concealed objects detection and segmentation more challenging. This paper proposes a two-stage segmentation scheme for active THz images. The first stage is rough segmentation, in which K- means clustering and region analysis are applied. After this, the segmentation result may still not be satisfactory due to the low quality of active THz image. Then, the active shape model (ASM) is introduced in the second fine segmentation stage. Considering the shape prior of the target object, ASM could further improve the segmentation result. Experimental results verified that the proposed scheme can segment the concealed objects in the active terahertz images accurately.

Background subtraction algorithms are widely utilized as a technology for segmentation of background and target objects in images. In particular, the simple background subtraction algorithm is used in many systems for its ease and low cost of implementation. However, because this algorithm relies only on the intensity difference, it has various problems, such as low tolerance for poor illumination and shadows and the inability to distinguish objects from their background when their intensities are similar. In an earlier study we proposed a new statistic, known as radial reach correlation (RRC), for distinguishing similar areas and dissimilar areas when comparing background images and target images at the pixel level. And we achieved a robust background subtraction by evaluating the local texture in images. In this study we extended this method further and developed a method to ensure stable background separation even in cases where the image texture is feeble and the intensity distribution is biased.

Segmentation and classification of objects in images is one of the most important and yet one of the most complex problems in computer vision. Many different things contribute to the complexity of this problem, but amongst those that are most significant are variations in object appearance due to various perspectives, changes in scene illumination and typically a large number of partially occluded objects in a scene. Humans have the ability to successfully deal with these complications and to recognize objects in scenes without seemingly much effort. However, to aid their recognition process, humans use different sources of information such as previous experiences, context of objects placed in a scene and rules learned in the past that suggest how the physical world should appear and behave. Computer vision is trying to replicate the same strategies used by human vision and apply them to computer systems with the aim of improving typical tasks of detection, localization and classification of image objects. In this work we propose a new model for natural image object classification using contextual information at the level of image segments. Context modeling is largely independent of appearance-based classification and proposed model enables simple upgrade of existing systems with information from global and/or local context. Context modeling is based on non-parametric use of appearance-based classification results which is a novel approach compared to previous systems that model context on a limited number of rules expressed with a fixed set of parameters. Model implementation resulted in a system that, in our simulations, showed stable improvement of the appearance-based object classification. DOI: http://dx.doi.org/10.5755/j01.itc.46.1.13610

In this paper, we propose a new approach for the detection and classification of rice diseases that combines the use of deep learning and level set methods. Our approach aims to improve the performance of rice disease classification and object detection by using an efficient image processing technique with deep learning. The proposed method will be evaluated based on Mean Average Precision (mAP), Accuracy, Precision, Recall, and F1-score. The results indicate that our proposed method improved the performance for most of the 16 classes of rice diseases and also it was able to improve performance even when using a small amount of labeled data. We also found that our proposed algorithm produced segmentation maps with significantly smaller computational time and the use of bias field estimation helped in the segmentation of complicated foreground objects in images by modifying unnecessary details. Overall, this research has the potential to be a valuable tool in the field of agriculture, where early and accurate detection of rice diseases is crucial for preventing the spread of the disease and maintaining crop yields.

This paper addresses the problem of recognizing and reconstructing real-world objects in cluttered environments to enable service robot to grasp the objects and manipulate them. A novel approach to combine disparity segmentation method with the closed-loop color region based segmentation is presented. Disparity map segmentation leads to definition of object region of interest (ROI) enabling autonomous functioning of robot system in cluttered environments. Closed-loop object region segmentation is robust against variable illumination providing reliable operation of robot system in different lighting conditions. Starting from the segmented object in both stereo images, the 3D contour of the object is generated and the object geometry is recovered from it. The proposed method needs no a-priori knowledge about the object color, its appearance or geometry. The performance of the presented method has been tested within the working scenario of the assistive robotic system FRIEND.

Segmentation of ultrasound images is challenging because of the noisy nature and subtle boundaries of objects in ultrasound images. This paper discusses object segmentation and identification for ultrasound fetal images. The feature space for segmentation consists of information extracted from three sources: gray level, texture, and wavelet-based decomposition. Several texture features, including Laws' texture-energy measures and features based on local gray level run-length, were found useful for segmentation. An unsupervised clustering procedure was used to classify each pixel into its most probable class. Morphological operations were used to remove noisy structures from the original gray level images and to improve the boundaries of the segmented objects. An algorithm was developed to locate objects of interest based on a multiscale implementation of an image transform. Fetal heads were identified and their corresponding measurements are made automatically. The method was tested with a set of clinical images. The resulting images showed clearly the segmented objects. The measurements agreed closely with a sonographer's measurements. The purposed method holds promise for processing and analyzing ultrasound fetal images.

The vision system provides an important way for construction robots to obtain the type and spatial location information of the object. The characteristics of the construction environment, construction object, and robot structure are jointly examined in this paper to propose an approach of object segmentation by spraying the robot based on multi-layer perceptron. Firstly, the hand-eye system experimental platform is built through establishing the mathematical model of the system and calibrating the parameters of the model. Secondly, effort is made to carry out research on image preprocessing algorithms and related experiments, and compare the effects of different binocular stereo-matching algorithms in the actual engineering environment. Finally, research and an experiment are conducted to identify the applicability and effect of the depth image object segmentation algorithm based on multi-layer perceptron. The experimental results prove that the application of multi-layer perceptron to object segmentation by spraying robots can meet the requirement on solution accuracy and is suitable for the object segmentation of complex projects in real life. This approach not only overcomes the shortcomings of the existing recognition methods that are poor in accuracy and difficult to be used widely, but also provides basic data for the subsequent three-dimensional reconstruction, thus making a significant contribution to the research of image processing by spraying robots.