This paper describes a color texture-based image segmentation system. The color texture informati... more This paper describes a color texture-based image segmentation system. The color texture information is obtained via modeling with the Multispectral Simultaneous Auto Regressive (MSAR) random field model. The general color content characterized by ratios of sample color means is also used. The image is segmented into regions of uniform color texture using an unsupervised histogram clustering approach that utilizes the
A large number of texture classification approaches have been developed in the past but most of t... more A large number of texture classification approaches have been developed in the past but most of these studies target gray-level textures. In this paper, supervised classification of color textures is considered. Several different Multispectral Random Field models are used to characterize the texture. The classifying features are based on the estimated parameters of these model and functions defined on them.
I wish to express my sincere gratitude and appreciation to my advisor, Dr. James G. Nagy, for his... more I wish to express my sincere gratitude and appreciation to my advisor, Dr. James G. Nagy, for his guidance and encouragement. I would also like to thank Dr. Ian Gladwell for advising and encouraging me to pursue this endeavor. I am thankful to Texas Instruments and ...
An adaptive line enhancer (ALE) is used to obtain estimates of the single sweep steady-state visu... more An adaptive line enhancer (ALE) is used to obtain estimates of the single sweep steady-state visual evoked potential (SSVEP). The method is seen to enhance the estimated signal-to-noise ratio of the single sweep SSVEP by as much as 10 dB.
A large number of texture classification approaches have been developed in the past but most of t... more A large number of texture classification approaches have been developed in the past but most of these studies target gray-level textures. In this paper, supervised classification of color textures is considered. Several different Multispectral Random Field models are used to characterize the texture. The classifying features are based on the estimated parameters of these model and functions defined on them.
In this paper, a robust, and fast system for recognition as well as pose estimation of a 3-D obje... more In this paper, a robust, and fast system for recognition as well as pose estimation of a 3-D object from a single 2-D perspective of it taken from an arbitrary viewpoint is developed. The approach is invariant to location, orientation, and scale of the object in the perspective. The silhouette of the object in the 2-D perspective is first normalized with respect to location and scale. A set of rotation invariant features derived from complex and orthogonal pseudo- Zernike moments of the image are then extracted. The next stage includes a bank of multilayer feed-forward neural networks (NN) each of which classifies the extracted features. The training set for these nets consists of perspective views of each object taken from several different viewing angles. The NNs in the bank differ in the size of their hidden layer nodes as well as their initial conditions but receive the same input. The classification decisions of all the nets are combined through a majority voting scheme. It is shown that this collective decision making yields better results compared to a single NN operating alone. After the object is classified, two of its pose parameters, namely elevation and aspect angles, are estimated by another module of NNs in a two-stage process. The first stage identifies the likely region of the space that the object is being viewed from. In the second stage, an NN estimator for the identified region is used to compute the pose angles. Extensive experimental studies involving clean and noisy images of seven military ground vehicles are carried out. The performance is compared to two other traditional methods, namely a nearest neighbor rule and a binary decision tree classifier and it is shown that our approach has major advantages over them.
A new quality measurement for video sequences utilized in video retrieval systems and visual data... more A new quality measurement for video sequences utilized in video retrieval systems and visual data mining applications is proposed. First, each frame of the sequence undergoes a segmentation step using extracted texture features from the gray-level cooccurrence matrix (GLCM) (Davis and Johns, 1979). Next, corresponding objects between adjacent frames are matched thus resulting in a 3-dimensional segmentation of the video into objects. Finally, color and texture features are extracted for each object in the sequence and provide the primary input in computing the quality measurement pertaining to the video. A low quality measurement may thus eliminate the possibility of the sequence being stored in a database retrieval system. The algorithm is tested on various types of video segments - pans, zooms, close-ups, and multiple objects' motion - with results included
A novel approach pertaining to the fast and efficient retrieval and storage of video sequences ut... more A novel approach pertaining to the fast and efficient retrieval and storage of video sequences utilizing MPEG-1/2 motion vectors is presented in this paper. A clip first must be segmented into consistent video segments based on the basic editing effects - cuts, dissolves, and wipes. All group of pictures (GOPs) are then extracted from the clip and decomposed further into I-frames and P-frames (B-frames are disregarded). The initial frame of the sequence is manually segmented into objects, and the selected objects are automatically tracked through the entire sequence using MPEG-1/2 motion vectors. Features pertaining to the edge histogram are extracted from each tracked object (I/P frames only) and then maintained with the associated frame. An example image or video clip is then presented to the system and the 5 best matching images are retrieved. Results are shown for the standard video sequences such as foreman, tennis, etc.
A new method for detecting the dissolve production effect within digital videos is proposed. Poss... more A new method for detecting the dissolve production effect within digital videos is proposed. Possible dissolve candidates are first identified based on the MPEG-7 edge histogram (Day, N. and Martinez, J.M., Proc. ISO/IEC/SC29/WG11 N4325, 2001) differences accumulated across a sampled region of the video. These potential candidates are then classified as a dissolve based on the analysis of tracking interesting objects within the considered video segment. The MPEG-7 descriptors (Day and Martinez, 2001), consisting of the edge histogram, homogenous texture, dominant colors and color structure (196 features in all), are then extracted from corresponding objects in successive frames for the duration of the potential dissolve sequence. The object's features are observed to undergo profound changes during a dissolve effect, while changing very little during other types of gradual transitions (e.g. camera panning and zooming). These object changes are used in classifying the sequence as a dissolve.
A new method for the temporal segmentation of video sequences into real-world objects is proposed... more A new method for the temporal segmentation of video sequences into real-world objects is proposed. First, each frame undergoes a color quantization step by matching like colors extracted from the previously processed frame. JSEG's color variance feature and texture features from the gray-level co-occurrence matrix (GLCM) are both extracted from each color-quantized frame and combined to obtain a more optimal image segmentation. Finally, a validation step is performed between the segmented regions of the currently processed frame and those in the previous frame, thus matching existing objects between frames and automatically detecting new objects upon their entrance into the scene. The new algorithm is tested on various video segments (pans, zooms, close-ups, and multiple-object motion) with results included
This paper describes a color texture-based image segmentation system. The color texture informati... more This paper describes a color texture-based image segmentation system. The color texture information is obtained via modeling with the Multispectral Simultaneous Auto Regressive (MSAR) random field model. The general color content characterized by ratios of sample color means is also used. The image is segmented into regions of uniform color texture using an unsupervised histogram clustering approach that utilizes the
A large number of texture classification approaches have been developed in the past but most of t... more A large number of texture classification approaches have been developed in the past but most of these studies target gray-level textures. In this paper, supervised classification of color textures is considered. Several different Multispectral Random Field models are used to characterize the texture. The classifying features are based on the estimated parameters of these model and functions defined on them.
I wish to express my sincere gratitude and appreciation to my advisor, Dr. James G. Nagy, for his... more I wish to express my sincere gratitude and appreciation to my advisor, Dr. James G. Nagy, for his guidance and encouragement. I would also like to thank Dr. Ian Gladwell for advising and encouraging me to pursue this endeavor. I am thankful to Texas Instruments and ...
An adaptive line enhancer (ALE) is used to obtain estimates of the single sweep steady-state visu... more An adaptive line enhancer (ALE) is used to obtain estimates of the single sweep steady-state visual evoked potential (SSVEP). The method is seen to enhance the estimated signal-to-noise ratio of the single sweep SSVEP by as much as 10 dB.
A large number of texture classification approaches have been developed in the past but most of t... more A large number of texture classification approaches have been developed in the past but most of these studies target gray-level textures. In this paper, supervised classification of color textures is considered. Several different Multispectral Random Field models are used to characterize the texture. The classifying features are based on the estimated parameters of these model and functions defined on them.
In this paper, a robust, and fast system for recognition as well as pose estimation of a 3-D obje... more In this paper, a robust, and fast system for recognition as well as pose estimation of a 3-D object from a single 2-D perspective of it taken from an arbitrary viewpoint is developed. The approach is invariant to location, orientation, and scale of the object in the perspective. The silhouette of the object in the 2-D perspective is first normalized with respect to location and scale. A set of rotation invariant features derived from complex and orthogonal pseudo- Zernike moments of the image are then extracted. The next stage includes a bank of multilayer feed-forward neural networks (NN) each of which classifies the extracted features. The training set for these nets consists of perspective views of each object taken from several different viewing angles. The NNs in the bank differ in the size of their hidden layer nodes as well as their initial conditions but receive the same input. The classification decisions of all the nets are combined through a majority voting scheme. It is shown that this collective decision making yields better results compared to a single NN operating alone. After the object is classified, two of its pose parameters, namely elevation and aspect angles, are estimated by another module of NNs in a two-stage process. The first stage identifies the likely region of the space that the object is being viewed from. In the second stage, an NN estimator for the identified region is used to compute the pose angles. Extensive experimental studies involving clean and noisy images of seven military ground vehicles are carried out. The performance is compared to two other traditional methods, namely a nearest neighbor rule and a binary decision tree classifier and it is shown that our approach has major advantages over them.
A new quality measurement for video sequences utilized in video retrieval systems and visual data... more A new quality measurement for video sequences utilized in video retrieval systems and visual data mining applications is proposed. First, each frame of the sequence undergoes a segmentation step using extracted texture features from the gray-level cooccurrence matrix (GLCM) (Davis and Johns, 1979). Next, corresponding objects between adjacent frames are matched thus resulting in a 3-dimensional segmentation of the video into objects. Finally, color and texture features are extracted for each object in the sequence and provide the primary input in computing the quality measurement pertaining to the video. A low quality measurement may thus eliminate the possibility of the sequence being stored in a database retrieval system. The algorithm is tested on various types of video segments - pans, zooms, close-ups, and multiple objects' motion - with results included
A novel approach pertaining to the fast and efficient retrieval and storage of video sequences ut... more A novel approach pertaining to the fast and efficient retrieval and storage of video sequences utilizing MPEG-1/2 motion vectors is presented in this paper. A clip first must be segmented into consistent video segments based on the basic editing effects - cuts, dissolves, and wipes. All group of pictures (GOPs) are then extracted from the clip and decomposed further into I-frames and P-frames (B-frames are disregarded). The initial frame of the sequence is manually segmented into objects, and the selected objects are automatically tracked through the entire sequence using MPEG-1/2 motion vectors. Features pertaining to the edge histogram are extracted from each tracked object (I/P frames only) and then maintained with the associated frame. An example image or video clip is then presented to the system and the 5 best matching images are retrieved. Results are shown for the standard video sequences such as foreman, tennis, etc.
A new method for detecting the dissolve production effect within digital videos is proposed. Poss... more A new method for detecting the dissolve production effect within digital videos is proposed. Possible dissolve candidates are first identified based on the MPEG-7 edge histogram (Day, N. and Martinez, J.M., Proc. ISO/IEC/SC29/WG11 N4325, 2001) differences accumulated across a sampled region of the video. These potential candidates are then classified as a dissolve based on the analysis of tracking interesting objects within the considered video segment. The MPEG-7 descriptors (Day and Martinez, 2001), consisting of the edge histogram, homogenous texture, dominant colors and color structure (196 features in all), are then extracted from corresponding objects in successive frames for the duration of the potential dissolve sequence. The object's features are observed to undergo profound changes during a dissolve effect, while changing very little during other types of gradual transitions (e.g. camera panning and zooming). These object changes are used in classifying the sequence as a dissolve.
A new method for the temporal segmentation of video sequences into real-world objects is proposed... more A new method for the temporal segmentation of video sequences into real-world objects is proposed. First, each frame undergoes a color quantization step by matching like colors extracted from the previously processed frame. JSEG's color variance feature and texture features from the gray-level co-occurrence matrix (GLCM) are both extracted from each color-quantized frame and combined to obtain a more optimal image segmentation. Finally, a validation step is performed between the segmented regions of the currently processed frame and those in the previous frame, thus matching existing objects between frames and automatically detecting new objects upon their entrance into the scene. The new algorithm is tested on various video segments (pans, zooms, close-ups, and multiple-object motion) with results included
Uploads