Showing papers by "Bogdan Smolka published in 2014"

Self-adaptive algorithm for segmenting skin regions

Abstract: In this paper, we introduce a new self-adaptive algorithm for segmenting human skin regions in color images Skin detection and segmentation is an active research topic, and many solutions have been proposed so far, especially concerning skin tone modeling in various color spaces Such models are used for pixel-based classification, but its accuracy is limited due to high variance and low specificity of human skin color In many works, skin model adaptation and spatial analysis were reported to improve the final segmentation outcome; however, little attention has been paid so far to the possibilities of combining these two improvement directions Our contribution lies in learning a local skin color model on the fly, which is subsequently applied to the image to determine the seeds for the spatial analysis Furthermore, we also take advantage of textural features for computing local propagation costs that are used in the distance transform The results of an extensive experimental study confirmed that the new method is highly competitive, especially for extracting the hand regions in color images

Advances in Low-Level Color Image Processing

Abstract: Color perception plays an important role in object recognition and scene understanding both for humans and intelligent vision systems. Recent advances in digital color imaging and computer hardware technology have led to an explosion in the use of color images in a variety of applications including medical imaging, content-based image retrieval, biometrics, watermarking, digital inpainting, remote sensing, visual quality inspection, among many others. As a result, automated processing and analysis of color images has become an active area of research, to which the large number of publications of the past two decades bears witness. The multivariate nature of color image data presents new challenges for researchers and practitioners as the numerous methods developed for single channel images are often not directly applicable to multichannel ones. The goal of this volume is to summarize the state-of-the-art in the early stages of the color image processing pipeline.

Eye blink detection using Local Binary Patterns

Abstract: In the paper a novel approach to the problem of eye blink detection in video sequences is proposed. The introduced method is utilizing the technique of Local Binary Patterns (LBP), which enables to build a descriptor capturing the features of the current eye state. In the initial step, the histogram of LBP describing the open eye is constructed and afterwards it serves as a template, which is compared with the histogram of the LBP of subsequent frames. The eye blinks are detected as sharp peaks of the dissimilarity between the template and the histogram of the current frame. The efficiency of the proposed eye blink detector has been compared with the state-of-the-art approaches using two video databases. The analysis of the results shows that the proposed technique effectiveness is superior to the existing methods of eye blink detection.

Application of the Extended Distance Transformation in digital image colorization

Abstract: In this paper we present a novel colorization scheme that takes advantage of the modified morphological distance transform to propagate the color, scribbled by a user on the grayscale image First, based on the scribbled image, the topological distance values are computed for each image pixel, describing its distance to the inserted color markers These values are then complemented with the structural information and luminance changes derived from the original grayscale image The distances are then used along with gradient based features to reproduce original image structures while propagating the new colors obtained during the additive color blending process Extensive experiments performed on various kinds of natural images demonstrated the effectiveness of the proposed colorization method They also showed that the main advantage of the presented algorithm is its computational speed and ability to produce visually pleasing colorization results promptly after providing the color information

Gaze direction estimation from static images

Abstract: This study presents a novel multilevel algorithm for gaze direction recognition from static images. Proposed solution consists of three stages: (i) eye pupil localization using a multistage ellipse detector combined with a support vector machines verifier, (ii) eye bounding box localization calculated using a hybrid projection function and (iii) gaze direction classification using support vector machines and random forests. The proposed method has been tested on Eye-Chimera database with very promising results. Extensive tests show that eye bounding box localization allows us to achieve highly accurate results both in terms of eye location and gaze direction classification.

Adaptive Non-local Means Filtering for Speckle Noise Reduction

Abstract: The aim of this study is to present the results of investigations concerning the evaluation of non-local means filter for multiplicative noise removal in ultrasonographic images. In this work a comparison of different techniques based on the concept of the non-local means filtering and a novel application for a filter called trimmed non-local means has been presented. The proposed modification is a generalization of the non-local means algorithm, in which the pixels are ordered using rank-ordered absolute differences statistic and only the most centrally located pixels in the filtering window are considered and used to calculate the weights needed for the averaging operation. The experiments confirmed that the proposed algorithm achieves comparable results with the existing state-of-the-art denoising schemes in suppressing multiplicative noise in ultrasound images.

Isoline Based Image Colorization

Abstract: Colorization is a process of adding colors to black and white images and videos. Each scalar pixel value is replaced by three values which describe, depending on color mode, e.g. The luminance and the chrominance or red-green-blue components. By introducing colors, the resulting image benefit greatly. Not only that it is more attractive for a viewer, but also the perception of hardly noticeable features increases, which is crucial e.g. In medical image analysis. In our paper we present a novel algorithm for grayscale image colorization. The idea is based on the concept of isolines on geographical maps. For each color indicated by a user with scribbles, we create distance maps which present the measure of distance between a pixel and the colored region. The resulting color is calculated as a weighted average of all colors indicated by the user in a form of scribbles. The method is computationally efficient and can be adjusted by several parameters to obtain optimal colorization results. We present several multimedia and biomedical images, colorized using our approach.

Robustified vector median filter

Abstract: In this paper a novel, efficient extension of the vector median filter intended for the suppression of impulsive noise in color images is proposed. The new filter operates on the trimmed distances between color pixels belonging to the filtering window. The cumulated distances calculated for each pixel in the local window is used to perform the reduced vector ordering, which allows to find the pixel which is centrally located in the cluster of most similar samples. The introduced generalization allows to improve the effectiveness of the standard vector median filter and can be used for more efficient restoration of color images distorted by high intensity impulsive noise. The unique property of the described filtering framework is its ability to sharpen the image edges which was quantified using a novel image restoration measure. Additionally, the proposed vector median extension does not increase its computational intensity, which allows to use it in real time applications.

Thresholded median filter for the impulsive noise removal in digital images

Abstract: In the paper a novel approach to the problem of impulsive noise removal in digital images is proposed. The new technique is based on the concepts derived from the robust statistics and can be regarded as a generalization of the median filter. The filter output is the pixel minimizing the trimmed cumulated sum of intensity distances to all other pixels belonging to the same filtering window. The performed experiments revealed that the new technique offers much better performance than the standard median filter. The unique feature of the new denoising scheme is its ability to sharpen image edges, while preserving its details.

Novel Approach to Noise Reduction in Ultrasound Images Based on Geodesic Paths

Abstract: In this paper a new method of multiplicative noise reduction in ultrasound images is proposed. The novel technique is a modification of the bilateral denosing scheme, which takes into account the similarity of pixels and their spatial distance. The filter output is calculated as a weighted average of the pixels which are in the neighborhood relation with the center of the filtering window, and the weights are functions of the minimal connection costs between surounding pixels. Experimental results show that the proposed method yields significantly better results than the other techniques in case of ultrasound images contaminated by medium and strong multiplicative noise disturbances.

Visualization enhancement of segmented images using genetic algorithm

Abstract: The aim of the paper is to present the results of investigations concerning the implementation of pseudocolor visualization algorithm of segmented images, capable to find the color combination producing maximum contrast between the segmented areas. Very often there is a need of visualization of segmentation results and usually they are presented by assigning colors randomly or from predefined palettes, what could decrease the visualization effect, when neighboring regions have assigned similar colors. To alleviate this problem, we propose novel methodology for deriving optimized visualization based on maximizing local distance between colors. In the paper we present visualization results using a new color contrast measure optimized with a genetic algorithm and compare the effectiveness with a greedy algorithm. The proposed method can be used to obtain visually pleasing pseudocolor encoded images of segmentation results which can be useful for the presentation of various kinds of visual information.

Optimal classification method for smiling vs neutral facial display recognition

Abstract: Human face depicts what happens in the soul, therefore correct recognition of emotion on the basis of facial display is of high importance. This work concentrates on the problem of optimal classification technique selection for solving the issue of smiling versus neutral face recognition. There are compared most frequently applied classification techniques: k-nearest neighbourhood, support vector machines, and template matching. Their performance is evaluated on facial images from several image datasets, but with similar image description methods based on local binary patterns. According to the experiments results the linear support vector machine gives the most satisfactory outcomes for all conditions.

Self-Adaptive Skin Segmentation in Color Images

Abstract: In this paper, we present a new method for skin detection and segmentation, relying on spatial analysis of skin-tone pixels. Our contribution lies in introducing self-adaptive seeds, from which the skin probability is propagated using the distance transform. The seeds are determined from a local skin color model that is learned on-line from a presented image, without requiring any additional information. This is in contrast to the existing methods that need a skin sample for the adaptation, e.g., acquired using a face detector. In our experimental study, we obtained F-score of over 0.85 for the ECU benchmark, and this is highly competitive compared with several state-of-the-art methods.

On the robustified median filter for the reduction of impulsive noise in digital images

Abstract: In this work an effective approach to the problem of impulsive noise removal in digital images is proposed. The novel filtering design is utilizing the elements of robust statistics and can be treated as a generalization of the standard median denoising algorithm. The output of the filter, operating in the local window, is defined as the pixel which minimizes the cumulated sum of trimmed absolute intensity differences between the pixels. Extensive experiments have revealed that the new technique yields much better performance when compared with the median filter. The beneficial feature of the new denoising framework is its ability to sharpen edges, while preserving image details. The simplicity of the proposed median extension enables its application in real time image enhancement.

Non-local Mean-Shift filter for the reduction of multiplicative noise in digital images

Abstract: In this paper a new method for the reduction of multiplicative noise in digital images is described. The proposed algorithm is a modification of the Mean-Shift (MS) filter which is based on the concept of the Non-Local Means (NLM) denoising. The proposed algorithm does not focus on single pixels only, as in the case of the mean-shift technique, but also on their neighborhoods. The performance of the novel approach is experimentally verified and the obtained results prove that the new design is superior both to the MS and NLM techniques.