Adilson Gonzaga received the doctoral degree in applied physics from the University of Sao Paulo, Sao Carlos, Brazil, in 1991. He is an Associate Professor at University of Sao Paulo. Since 1973 he has been working at School of Engineering of São Carlos, Electrical Engineering Department, where he has done research in the field of Computer Vision, Image Processing and Biometry. He is the supervisor of graduate students working with Computer Vision. His current interest is Automated Visual Inspection, Biometry, Robot Vision, Perception and Artificial Intelligence. He is the coordinator of the Computer Vision Laboratory (LAVI).
This paper deals with texture segmentation in images of wooden plates. These plates are used in p... more This paper deals with texture segmentation in images of wooden plates. These plates are used in pencil industry and its quality reflects on the quality of the industrialized product. We have used the spatial gray level dependence method for analyzing the plate surface, detecting nodes, stripes, and defects. This statistical method has many applications in several fields with textures. We classify the pates based on the industry requirements. With images of 128 X 256 pixels and 256 gray levels, we have used a 64 X 64 window placed over 8 non- overlapping areas. For each position the window computes the co-occurrence matrices for the directions 0, 45, 90 and 135 degrees with a chessboard distance equal to 1. The values of he angular second moment average, computed in the four directions versus the plate area, are plotted in a feature space. This shows clusters of similar textures around constant values. The good plates have a homogeneous visual appearance. The unusable plates have nodes, stripes and defects. Between then exist the high stripped plates and the medium stripped plates. We can classify then in good one or unusable depending on the industry demand. The method shows a good efficiency for discriminating between textures of the same material but with a random distribution established by the nature.
Indirect Immunofluorescence Images (IIF) are generated based on a biological response to a specif... more Indirect Immunofluorescence Images (IIF) are generated based on a biological response to a specific light spectrum. The analysis of this type of image is helpful for the diagnosis of systemic autoimmune diseases. The Antinuclear Antibodies (ANAs) are detected, in the patient serum, by a specific fluorescence pattern on Human Epithelial type 2 (HEp-2) cells, due to its large range of antigens. In this paper, the two rotation invariant texture descriptors, Sampled Local Mapped Pattern Magnitude (SLMP_M) and Completed Local Binary Pattern Sign and Magnitude (CLBP_S/M) with a multiresolution configuration are investigated to classify 1455 Human Epithelial type 2 cells, extracted from 28 specimen images divided into six pattern categories (centromere, coarse speckled, cytoplasmatic, fine speckled, homogeneous and nucleolar). Results have shown that the multiresolution configuration improves the classification performance and that the SLMP_M descriptor is more suitable and faster to describe this type of images.
Abstract. This paper presents a color–texture descriptor based on the local mapped pattern approa... more Abstract. This paper presents a color–texture descriptor based on the local mapped pattern approach for color–texture classification under different lighting conditions. The proposed descriptor, namely extended color local mapped pattern (ECLMP), considers the magnitude of the color vectors inside the RGB cube to extract color–texture information from the images. These features are combined with texture information from the luminance image in a multiresolution fashion to get the ECLMP feature vector. The robustness of the proposed method is evaluated using the RawFooT, KTH-TIPS-2b, and USPtex databases. The experimental results show that the proposed descriptor is more robust to changes in the illumination condition than 22 alternative commonly used descriptors.
A feature descriptor aims at the characterization of images whose discriminative data are used by... more A feature descriptor aims at the characterization of images whose discriminative data are used by different applications, as image recognition and classification. An efficient feature descriptor can accurately describe the content, and maintain low computational resources. This paper proposes a new feature descriptor, called Modified Center-Symmetric Local Mapped Pattern (MCS-LMP), which is based on the Center-Symmetric Local Mapped Pattern (CS-LMP) descriptor. By using the central pixel in the calculations, we can better characterize an image content, while using a reduced descriptor size. The descriptor was evaluated in an object classification task through the comparison of the results with those of CS-LMP, SIFT and LIOP. It could improve the sensibility of classification at shorter computational time.
Face recognition has received significant attention during the past several years. It is a challe... more Face recognition has received significant attention during the past several years. It is a challenge task because faces can be affected by scale, noises, face expression, illumination, color or pose variations. The most robust methodologies related to these variations are based on "key points?" localization, followed by the application of a local descriptor to each surrounding region. Such descriptors are associated to clustering algorithms or histogram representation based on Bag of Features (BoF). In the BoF approach, the codebook can effectively describe objects by their appearance based on local texture. Based on texture descriptors proposed previously for image detection, we propose in this paper the application of such descriptors for face recognition. We evaluate the performance of our methodology using Feret, ORL and Yale databases, comparing our descriptor against SIFT and LIOP descriptors, and also other methodologies recently published in the literature.
The biometric periocular trait refers to the face region in the vicinity of the eye, including th... more The biometric periocular trait refers to the face region in the vicinity of the eye, including the eyelids, eyelashes and eyebrows. The periocular region has emerged as a promising trait for unconstrained biometrics, due to recent advances of convolutional neural networks and the demand for robust face or iris recognition systems. The periocular region can offer global information about the eye shape, and about the texture of the iris, sclera and skin around the eyes. However, periocular biometrics is a relatively new area of research. Thus, it's important to understand the uniqueness and stability of this trait, taking into account the best accuracies obtained by deep learning methods applied on biometric image recognition. In this work, we investigate if changes in the periocular region, caused by facial expressions, affect the recognition accuracy. We apply an existing pretrained CNN architecture, called MobileNet, to the task of periocular recognition. The periocular images used in the experiments were extracted from the Extended Cohn-Kanade expression database. The best results were obtained when the network was tested with similar samples to those contained in the training set. We concluded that the CNN is sensitive to changes in the facial expressions and samples of all expressions are required for training aiming the best accuracy.
Medical Imaging 2018: Computer-Aided Diagnosis, 2018
This paper proposes a method to reduce the number of false-positives (FP) in a computer-aided det... more This paper proposes a method to reduce the number of false-positives (FP) in a computer-aided detection (CAD) scheme for automated detection of architectural distortion (AD) in digital mammography. AD is a subtle contraction of breast parenchyma that may represent an early sign of breast cancer. Due to its subtlety and variability, AD is more difficult to detect compared to microcalcifications and masses, and is commonly found in retrospective evaluations of false-negative mammograms. Several computer-based systems have been proposed for automated detection of AD in breast images. The usual approach is automatically detect possible sites of AD in a mammographic image (segmentation step) and then use a classifier to eliminate the false-positives and identify the suspicious regions (classification step). This paper focus on the optimization of the segmentation step to reduce the number of FPs that is used as input to the classifier. The proposal is to use statistical measurements to score the segmented regions and then apply a threshold to select a small quantity of regions that should be submitted to the classification step, improving the detection performance of a CAD scheme. We evaluated 12 image features to score and select suspicious regions of 74 clinical Full-Field Digital Mammography (FFDM). All images in this dataset contained at least one region with AD previously marked by an expert radiologist. The results showed that the proposed method can reduce the false positives of the segmentation step of the CAD scheme from 43.4 false positives (FP) per image to 34.5 FP per image, without increasing the number of false negatives.
This paper presents a new local micro-pattern texture descriptor for the detection of Architectur... more This paper presents a new local micro-pattern texture descriptor for the detection of Architectural Distortion (AD) in digital mammography images. AD is a subtle contraction of breast parenchyma that may represent an early sign of breast cancer. Due to its subtlety and variability, AD is more difficult to detect compared to microcalcifications and masses, and is commonly found in retrospective evaluations of false-negative mammograms. Several computer-based systems have been proposed for automatic detection of AD, but their performance are still unsatisfactory. The proposed descriptor, Local Mapped Pattern (LMP), is a generalization of the Local Binary Pattern (LBP), which is considered one of the most powerful feature descriptor for texture classification in digital images. Compared to LBP, the LMP descriptor captures more effectively the minor differences between the local image pixels. Moreover, LMP is a parametric model which can be optimized for the desired application. In our work, the LMP performance was compared to the LBP and four Haralick's texture descriptors for the classification of 400 regions of interest (ROIs) extracted from clinical mammograms. ROIs were selected and divided into four classes: AD, normal tissue, microcalcifications and masses. Feature vectors were used as input to a multilayer perceptron neural network, with a single hidden layer. Results showed that LMP is a good descriptor to distinguish AD from other anomalies in digital mammography. LMP performance was slightly better than the LBP and comparable to Haralick's descriptors (mean classification accuracy = 83%).
This work describes a technique of classification of remote sensing images using the texture attr... more This work describes a technique of classification of remote sensing images using the texture attribute as source of data. The processes of classification/segmentation are described and the use of Gabor Filters for the supervised classification of such images is proposed here, for the purpose of separating areas of water, vegetation and urban region. For an image characteristic extraction was chosen an spectrum approach, which try to identify a periodicity in the image, through an identification of peaks of high energy in the ...
This paper deals with texture segmentation in images of wooden plates. These plates are used in p... more This paper deals with texture segmentation in images of wooden plates. These plates are used in pencil industry and its quality reflects on the quality of the industrialized product. We have used the spatial gray level dependence method for analyzing the plate surface, detecting nodes, stripes, and defects. This statistical method has many applications in several fields with textures. We classify the pates based on the industry requirements. With images of 128 X 256 pixels and 256 gray levels, we have used a 64 X 64 window placed over 8 non- overlapping areas. For each position the window computes the co-occurrence matrices for the directions 0, 45, 90 and 135 degrees with a chessboard distance equal to 1. The values of he angular second moment average, computed in the four directions versus the plate area, are plotted in a feature space. This shows clusters of similar textures around constant values. The good plates have a homogeneous visual appearance. The unusable plates have nodes, stripes and defects. Between then exist the high stripped plates and the medium stripped plates. We can classify then in good one or unusable depending on the industry demand. The method shows a good efficiency for discriminating between textures of the same material but with a random distribution established by the nature.
Indirect Immunofluorescence Images (IIF) are generated based on a biological response to a specif... more Indirect Immunofluorescence Images (IIF) are generated based on a biological response to a specific light spectrum. The analysis of this type of image is helpful for the diagnosis of systemic autoimmune diseases. The Antinuclear Antibodies (ANAs) are detected, in the patient serum, by a specific fluorescence pattern on Human Epithelial type 2 (HEp-2) cells, due to its large range of antigens. In this paper, the two rotation invariant texture descriptors, Sampled Local Mapped Pattern Magnitude (SLMP_M) and Completed Local Binary Pattern Sign and Magnitude (CLBP_S/M) with a multiresolution configuration are investigated to classify 1455 Human Epithelial type 2 cells, extracted from 28 specimen images divided into six pattern categories (centromere, coarse speckled, cytoplasmatic, fine speckled, homogeneous and nucleolar). Results have shown that the multiresolution configuration improves the classification performance and that the SLMP_M descriptor is more suitable and faster to describe this type of images.
Abstract. This paper presents a color–texture descriptor based on the local mapped pattern approa... more Abstract. This paper presents a color–texture descriptor based on the local mapped pattern approach for color–texture classification under different lighting conditions. The proposed descriptor, namely extended color local mapped pattern (ECLMP), considers the magnitude of the color vectors inside the RGB cube to extract color–texture information from the images. These features are combined with texture information from the luminance image in a multiresolution fashion to get the ECLMP feature vector. The robustness of the proposed method is evaluated using the RawFooT, KTH-TIPS-2b, and USPtex databases. The experimental results show that the proposed descriptor is more robust to changes in the illumination condition than 22 alternative commonly used descriptors.
A feature descriptor aims at the characterization of images whose discriminative data are used by... more A feature descriptor aims at the characterization of images whose discriminative data are used by different applications, as image recognition and classification. An efficient feature descriptor can accurately describe the content, and maintain low computational resources. This paper proposes a new feature descriptor, called Modified Center-Symmetric Local Mapped Pattern (MCS-LMP), which is based on the Center-Symmetric Local Mapped Pattern (CS-LMP) descriptor. By using the central pixel in the calculations, we can better characterize an image content, while using a reduced descriptor size. The descriptor was evaluated in an object classification task through the comparison of the results with those of CS-LMP, SIFT and LIOP. It could improve the sensibility of classification at shorter computational time.
Face recognition has received significant attention during the past several years. It is a challe... more Face recognition has received significant attention during the past several years. It is a challenge task because faces can be affected by scale, noises, face expression, illumination, color or pose variations. The most robust methodologies related to these variations are based on "key points?" localization, followed by the application of a local descriptor to each surrounding region. Such descriptors are associated to clustering algorithms or histogram representation based on Bag of Features (BoF). In the BoF approach, the codebook can effectively describe objects by their appearance based on local texture. Based on texture descriptors proposed previously for image detection, we propose in this paper the application of such descriptors for face recognition. We evaluate the performance of our methodology using Feret, ORL and Yale databases, comparing our descriptor against SIFT and LIOP descriptors, and also other methodologies recently published in the literature.
The biometric periocular trait refers to the face region in the vicinity of the eye, including th... more The biometric periocular trait refers to the face region in the vicinity of the eye, including the eyelids, eyelashes and eyebrows. The periocular region has emerged as a promising trait for unconstrained biometrics, due to recent advances of convolutional neural networks and the demand for robust face or iris recognition systems. The periocular region can offer global information about the eye shape, and about the texture of the iris, sclera and skin around the eyes. However, periocular biometrics is a relatively new area of research. Thus, it's important to understand the uniqueness and stability of this trait, taking into account the best accuracies obtained by deep learning methods applied on biometric image recognition. In this work, we investigate if changes in the periocular region, caused by facial expressions, affect the recognition accuracy. We apply an existing pretrained CNN architecture, called MobileNet, to the task of periocular recognition. The periocular images used in the experiments were extracted from the Extended Cohn-Kanade expression database. The best results were obtained when the network was tested with similar samples to those contained in the training set. We concluded that the CNN is sensitive to changes in the facial expressions and samples of all expressions are required for training aiming the best accuracy.
Medical Imaging 2018: Computer-Aided Diagnosis, 2018
This paper proposes a method to reduce the number of false-positives (FP) in a computer-aided det... more This paper proposes a method to reduce the number of false-positives (FP) in a computer-aided detection (CAD) scheme for automated detection of architectural distortion (AD) in digital mammography. AD is a subtle contraction of breast parenchyma that may represent an early sign of breast cancer. Due to its subtlety and variability, AD is more difficult to detect compared to microcalcifications and masses, and is commonly found in retrospective evaluations of false-negative mammograms. Several computer-based systems have been proposed for automated detection of AD in breast images. The usual approach is automatically detect possible sites of AD in a mammographic image (segmentation step) and then use a classifier to eliminate the false-positives and identify the suspicious regions (classification step). This paper focus on the optimization of the segmentation step to reduce the number of FPs that is used as input to the classifier. The proposal is to use statistical measurements to score the segmented regions and then apply a threshold to select a small quantity of regions that should be submitted to the classification step, improving the detection performance of a CAD scheme. We evaluated 12 image features to score and select suspicious regions of 74 clinical Full-Field Digital Mammography (FFDM). All images in this dataset contained at least one region with AD previously marked by an expert radiologist. The results showed that the proposed method can reduce the false positives of the segmentation step of the CAD scheme from 43.4 false positives (FP) per image to 34.5 FP per image, without increasing the number of false negatives.
This paper presents a new local micro-pattern texture descriptor for the detection of Architectur... more This paper presents a new local micro-pattern texture descriptor for the detection of Architectural Distortion (AD) in digital mammography images. AD is a subtle contraction of breast parenchyma that may represent an early sign of breast cancer. Due to its subtlety and variability, AD is more difficult to detect compared to microcalcifications and masses, and is commonly found in retrospective evaluations of false-negative mammograms. Several computer-based systems have been proposed for automatic detection of AD, but their performance are still unsatisfactory. The proposed descriptor, Local Mapped Pattern (LMP), is a generalization of the Local Binary Pattern (LBP), which is considered one of the most powerful feature descriptor for texture classification in digital images. Compared to LBP, the LMP descriptor captures more effectively the minor differences between the local image pixels. Moreover, LMP is a parametric model which can be optimized for the desired application. In our work, the LMP performance was compared to the LBP and four Haralick's texture descriptors for the classification of 400 regions of interest (ROIs) extracted from clinical mammograms. ROIs were selected and divided into four classes: AD, normal tissue, microcalcifications and masses. Feature vectors were used as input to a multilayer perceptron neural network, with a single hidden layer. Results showed that LMP is a good descriptor to distinguish AD from other anomalies in digital mammography. LMP performance was slightly better than the LBP and comparable to Haralick's descriptors (mean classification accuracy = 83%).
This work describes a technique of classification of remote sensing images using the texture attr... more This work describes a technique of classification of remote sensing images using the texture attribute as source of data. The processes of classification/segmentation are described and the use of Gabor Filters for the supervised classification of such images is proposed here, for the purpose of separating areas of water, vegetation and urban region. For an image characteristic extraction was chosen an spectrum approach, which try to identify a periodicity in the image, through an identification of peaks of high energy in the ...
Uploads