Frédéric Comby

Variable structuring element based fuzzy morphological operations for single viewpoint omnidirectional images

Morphological tools can provide transformations suitable for real projective images, but the camera and objects to be analyzed have to be positioned in such a manner that a regular mesh on the objects projects a regular mesh on the image. A morphological modification of the image is thus the projection of an equivalent operation on the object. Otherwise, due to perspective effects, a morphological operation on the image is not the projection of an equivalent operation on the objects to be analyzed. With catadioptric omnidirectional images, it is almost impossible to place the sensor such that a regular mesh on the scene projects a regular mesh on the image. Nevertheless, with proper calibration of a central catadioptric system, the projection of a regular structuring element in a scene can be determined for each point on the image. The aim of this paper is to present new morphological operators that use this projective property. These operators make use of a structuring element of varying shape. Since this varying shape cannot be represented as a binary union of pixels, we propose to use a fuzzy extension of the classical gray-level morphology to account for this phenomenon. This fuzzy extension is performed via fuzzy integrals.

Fuzzy edge detection for omnidirectional images

The use of omnidirectional vision has increased during these past years. It provides a very large field of view. Nevertheless, omnidirectional images contain significant radial distortions and conventional image processing is not adapted to these specific images. This paper presents an edge detector adapted to the image geometry. Fuzzy sets will be used to take into account all imprecisions introduced by the sampling process. The Prewitt filter applied to omnidirectional image will be studied to illustrate this paper.

Camera model identification with the use of deep convolutional neural networks

In this paper, we propose a camera model identification method based on deep convolutional neural networks (CNNs). Unlike traditional methods, CNNs can automatically and simultaneously extract features and learn to classify during the learning process. A layer of preprocessing is added to the CNN model, and consists of a high pass filter which is applied to the input image. Before feeding the CNN, we examined the CNN model with two types of residuals. The convolution and classification are then processed inside the network. The CNN outputs an identification score for each camera model. Experimental comparison with a classical two steps machine learning approach shows that the proposed method can achieve significant detection performance. The well known object recognition CNN models, AlexNet and GoogleNet, are also examined.

Rough histograms for robust statistics

Applied statistics are widely used in pattern recognition and other computing applications to find the most likely value of a parameter. The use of classical empirical statistics is based upon assumption about normality of underlying density distribution of data. When the data is corrupted by contaminated noise, then classical tools are usually not robust enough and the estimation of the mode is biased. In this article, we propose to estimate the main mode of a distribution by means of a rough histogram and we show that this estimation is robust to contamination.

Fuzzy morphology for omnidirectional images

This paper describes morphological tools that have been adapted to omnidirectional catadioptric images.

Non-Additive Approach for Gradient-Based Edge Detection

In this paper, we propose a new method to perform the first derivative estimation of a discrete intensity distribution. This approach is based on a non-additive aggregation process and provides an estimate of the gradient as intervals instead of single values. These intervals are used to threshold a gradient-based edge detection and therefore discard spurious detections due to noise.

Camera model identification based machine learning approach with high order statistics features

Source camera identification methods aim at identifying the camera used to capture an image. In this paper we developed a method for digital camera model identification by extracting three sets of features in a machine learning scheme. These features are the co-occurrences matrix, some features related to CFA interpolation arrangement, and conditional probability statistics. These features give high order statistics which supplement and enhance the identification rate. The method is implemented with 14 camera models from Dresden database with multi class SVM classifier. A comparison is performed between our method and a camera fingerprint correlation-based method which only depends on PRNU extraction. The experiments prove the strength of our proposition since it achieves higher accuracy than the correlation-based method.

Highly specific pose estimation with a catadioptric omnidirectional camera

This article presents a new method for estimating the pose of para-catadioptric vision systems. It is based on the estimation of vanishing points associated with vertical edges of the environment. However, unlike classical approaches no feature (line, circle) extraction and/or identification is needed. A sampled domain of possible vanishing points is tested and histograms are build to characterize the soundness of these points. A specificity index allows to find the more relevant histogram and the pose of the sensor. This method has been tested on simulated and real images giving very promising results (maximum angular error of 0.18 degree).

Non-additive Approach for Omnidirectional Image Gradient Estimation

The way catadioptric images are acquired implies that they present radial distortions. Therefore, classical processing may not be suitable. This statement will be illustrated by considering edge detection matter. Classical edge detectors usually consist in three steps : gradient computation, maximization and thresholding. The two lasts steps use pixels neighborhood concept. On the opposite of perspective images where pixel neighborhood is intuitive, catadioptric images present radial resolution changes. Then, the size and shape of pixel neighborhood have to be depending on pixel location. This article presents a new gradient estimation approach based on non-additive kernels. This technique is adapted to catadioptric images and also provides a natural threshold discarding the arbitrary thresholding step.

Possibility Theory and Rough Histograms for Motion Estimation in a Video Sequence

This article proposes to use both theories of possibility and rough histograms to deal with estimation of the movement between two images in a video sequence. A fuzzy modeling of data and a reasoning based on imprecise statistics allow us to partly cope with the constraints associated to classical movement estimation methods such as correlation or optical flow based-methods. The theoretical aspect of our method will be explained in details, and its properties will be shown. An illustrative example will also be presented.