Enrique Alegre

Class distribution estimation based on the Hellinger distance

Class distribution estimation (quantification) plays an important role in many practical classification problems. Firstly, it is important in order to adapt the classifier to the operational conditions when they differ from those assumed in learning. Additionally, there are some real domains where the quantification task is itself valuable due to the high variability of the class prior probabilities. Our novel quantification approach for two-class problems is based on distributional divergence measures. The mismatch between the test data distribution and validation distributions generated in a fully controlled way is measured by the Hellinger distance in order to estimate the prior probability that minimizes this divergence. Experimental results on several binary classification problems show the benefits of this approach when compared to such approaches as counting the predicted class labels and other methods based on the classifier confusion matrix or on posterior probability estimations. We also illustrate these techniques as well as their robustness against the base classifier performance (a neural network) with a boar semen quality control setting. Empirical results show that the quantification can be conducted with a mean absolute error lower than 0.008, which seems very promising in this field.

Texture and moments-based classification of the acrosome integrity of boar spermatozoa images

The automated assessment of the sperm quality is an important challenge in the veterinary field. In this paper, we explore how to describe the acrosomes of boar spermatozoa using image analysis so that they can be automatically categorized as intact or damaged. Our proposal aims at characterizing the acrosomes by means of texture features. The texture is described using first order statistics and features derived from the co-occurrence matrix of the image, both computed from the original image and from the coefficients yielded by the Discrete Wavelet Transform. Texture descriptors are evaluated and compared with moments-based descriptors in terms of the classification accuracy they provide. Experimental results with a Multilayer Perceptron and the k-Nearest Neighbours classifiers show that texture descriptors outperform moment-based descriptors, reaching an accuracy of 94.93%, which makes this approach very attractive for the veterinarian community.

Statistical Approach to Boar Semen Evaluation Using Intracellular Intensity Distribution of Head Images

We propose a method for the classification of boar sperm heads based on their intracellular intensity distributions observed in microscopic images. The image pre-processing comprises segmentation of cell heads and normalization for brightness, contrast and size. Next, we define a model distribution of head intracellular intensity of an alive cell using a set of head images assumed to be alive by veterinary experts. We now consider two other sets of cell head images, one formed by heads assumed to be alive by experts and another formed by cells which present some abnormalities in their cytoplasm densities and are considered as dead by the experts. We define a measure of deviation from the model intensity distribution and for each head image of the two test sets we compute the deviation from the model. While the distributions of deviation values for alive and dead cells overlap, it is possible to choose an optimal value of a decision criterion for single cell classification in such a way that the error made in the estimation of the fraction of alive cells in a sample is minimal. In the range [0.7,1.0] of interesting values of the fraction of alive cells, the standard deviation of the fraction estimation error for samples of 100 head images is smaller than 0.04. Thus, in 95% of the cases the value of the fraction of alive cells in a sample estimated by a veterinary expert will be within 8% of the estimation made according to the proposed method. This result satisfies the requirements of veterinary practice.

Automatic classification of the acrosome status of boar spermatozoa using digital image processing and LVQ

We consider images of boar spermatozoa obtained with an optical phase-contrast microscope. Our goal is to automatically classify single sperm cells as acrosome-intact (class 1) or acrosome-damaged (class 2). Such classification is important for the estimation of the fertilization potential of a sperm sample for artificial insemination. We segment the sperm heads and compute a feature vector for each head. As a feature vector we use the gradient magnitude along the contour of the sperm head. We apply learning vector quantization (LVQ) to the feature vectors obtained for 320 heads that were labelled as intact or damaged using stains. A LVQ system with four prototypes (two for each class) allows us to classify cells with an overall test error of 6.8%. This is considered to be sufficient for semen quality control in an artificial insemination center.

Local Oriented Statistics Information Booster (LOSIB) for Texture Classification

Local oriented statistical information booster (LOSIB) is a descriptor enhancer based on the extraction of the gray level differences along several orientations. Specifically, the mean of the differences along particular orientations is considered. In this paper we have carried out some experiments using several classical texture descriptors to show that classification results are better when they are combined with LOSIB, than without it. Both parametric and non-parametric classifiers, Support Vector Machine and k-Nearest Neighbourhoods respectively, were applied to assess this new method. Furthermore, two different texture dataset were evaluated: KTH-Tips-2a and Brodatz32 to prove the robustness of LOSIB. Global descriptors such as WCF4 (Wavelet Co-occurrence Features), that extracts Haralick features from the Wavelet Transform, have been combined with LOSIB obtaining an improvement of 16.94% on KTH and 7.55% on Brodatz when classifying with SVM. Moreover, LOSIB was used together with state-of-the-art local descriptors such as LBP (Local Binary Pattern) and several of its recent variants. Combined with CLBP (Complete LBP), the LOSIB booster results were improved in 5.80% on KTH-Tips 2a and 7.09% on the Brodatz dataset. For all the tested descriptors, we have observed that a higher performance has been achieved, with the two classifiers on both datasets, when using some LOSIB settings.

Design of a Computer Vision System to Estimate Tool Wearing

Wear level of tool inserts in automated processes is tried using techniques of artificial vision. An application has been developed in Matlab that allows the acquisition of images with different resolutions and later on to process them. It is explained how the vision system used has been designed and implemented. The method for acquiring tool insert images and their treatment in the pre-processing, segmentation and post-processing is commented. First results are also presented using diverse texture descriptors. These first results must be corroborated carrying out new experiments with a bigger number of images.

Statistical approach to boar semen head classification based on intracellular intensity distribution

We propose a technique to compute the fraction of boar spermatozoid heads which present an intracellular density distribution pattern hypothesized as normal by veterinary experts. This approach offers a potential for digital image processing estimation of sperm capacitation which can substitute expensive staining techniques. We extract a model distribution from a training set of heads assumed as normal by veterinary experts. We also consider two other training sets, one with heads similar to the normal pattern and another formed by heads that substantially deviate from that pattern. For each spermatozoid head, a deviation from the model distribution is computed. This produces a conditional probability distribution of that deviation for each set. Using a set of test images, we determine the fraction of normal heads in each image and compare it with the result of expert classification. This yields an absolute error below 0.25 in the 89% of the samples.

Classification and Quantification Based on Image Analysis for Sperm Samples with Uncertain Damaged/Intact Cell Proportions

Classifying damaged-intact cells in a semen sample presents the peculiarity that the test class distribution is unknown. This paper studies under which design conditions the misclassification rate is minimum for the uncertainty region of interest (ratio of damaged cells lower than 20%) and (b) deals with quantifying the proportion of damaged/intact sperm cells in a given sample based on computer vision and supervised learning. We have applied a discrete wavelet transform to the spermatozoa head images and computed the mean and standard deviation (WSF) and four Haralick descriptors (WCF). Using a backpropagation neural network, the error rate averaged over distributions in the region of interest is 4.85% with WCF. The assessment of several quantification methods shows the conditions under which the Adjusted Count method leads to an overall mean absolute error of 3.2 and the Classify & Count method yields 2.4, both with WCF features. Deviations of this order are considered reasonable for this field.

Adaptive local binary pattern with oriented standard deviation (ALBPS) for texture classification

AbstractA new method to describe texture images using a hybrid combination of local and global texture descriptors is proposed in this paper. In this regard, a new adaptive local binary pattern (ALBP) descriptor is presented in order to carry out the local description. It is built by adding oriented standard deviation information to an ALBP descriptor in order to achieve a more complete representation of the images, and hence, it has been called adaptive local binary pattern with oriented standard deviation (ALBPS). Regarding semen vitality assessment, ALBPS outperformed previous literature works with an 81.88% accuracy and also yielded higher hit rates than the LBP and ALBP baseline methods. Concerning the global description of the images, several classical texture algorithms were tested and a descriptor based on wavelet transform and Haralick feature extraction (wavelet concurrent feature 13 (WCF13)) obtained the best results. Both local and global descriptors were combined, and the classification was carried out with a support vector machine. Two data sets have been evaluated: textures under varying illumination, pose and scale (KTH-TIPS) 2a data set and a second spermatozoa boar data set used to distinguish between dead or alive sperm heads. Therefore, our proposal is novel in three ways. First, a new local feature extraction method ALBPS is introduced. Second, a hybrid method combining the proposed local ALBPS and a global descriptor is presented, outperforming our first approach and all other methods evaluated for this problem. Third, texture classification accuracy is greatly improved with the two former texture descriptors presented. F score and accuracy values were computed in order to measure the performance. The best overall result was yielded by combining ALBPS with WCF13, reaching an F score = 0.886 and an accuracy of 85.63% in the spermatozoa data set and an 84.45% of hit rate in the KTH-TIPS 2a.

Computer Vision and Classification Techniques on the Surface Finish Control in Machining Processes

This work presents a method to perform a surface finish control using a computer vision system. Test parts used were made of AISI 303 stainless steel and were machined with a MUPEM CNC multi-turret parallel lathe. Using a Pulnix PE2015 B/W camera, a diffuse illumination and a industrial zoom, 140 images were acquired. We have applied a vertical Prewitt filter to all the images obtaining two sets, the original one and the filtered. We have described the images using three different methods. The first features vector was composed by the mean, standard deviation, skewness and kurtosis of the image histogram. The second features vector was made up by four Haralick descriptors --- contrast, correlation, energy and homogeneity. The last one was composed by 9 Laws descriptors. Using k-nn we have obtained a hit rate around 90 % with filtered images and, the best one, using Laws features vector of 92.14% with unfiltered images. These results show that it is feasible to use texture descriptors to evaluate the rugosity of metallic parts in the context of product quality inspection.