Bozidar Potocnik

Automated computer-assisted detection of follicles in ultrasound images of ovary

Monitoring the follicles in womens ovaries is especially important in human reproduction. Today, the monitoring of follicles is done with human interaction. Such monitoring can be very demanding and inaccurate, and in most cases signifies additional burdens for the experts. In this paper, a new algorithm for automated computer-assisted detection of follicles in the ultrasound images of the ovary is proposed. It has a typical object recognition scheme (preprocessing, segmentation, and classification). The algorithm is based on the following idea: first, the ovary is estimated (coarsely) and then follicles are searched for. The methods used are known from literature (despeckle filter, Kirschs operator, optimal thresholding, thinning, shape descriptions, classification), and the majority of our work was done experimenting with these methods and selecting the appropriate thresholds. The algorithms computational complexity is of order of O(n2), which means about 6 min of processing time per an ultrasound image of dimensions of 768 × 576 pixels on HP 715 machines. It has been tested on a set of 20 real ultrasound images of the ovary. The recognition rate of follicles with these procedures was around 62%. The algorithm is not perfect, but it will be further modified and improved, as indicated in our conclusions.

Assessment of Region-Based Moment Invariants for Object Recognition

Geometric region-based moments as features for invariant object recognition are studied. Theoretically rotation, translation, and scale invariant Hu, Zernike, and Krawtchouk moments are used as features for region description. Accuracy of such description and efficiency is tested by recognition of letters and digits from extended Slovenian alphabet. Ten testing samples in six different image resolutions are constructed for each character from learning set. Testing set consists of 390 samples per resolution (altogether 2340 samples). Recognition accuracy obtained by using Hu moments is 95.6%, 87.4% with Zernike moments, and with Krawtchouk moments 64.1%. Object recognition by using Krawtchouk moments is the most sensitive to object rotation and scaling, which is confirmed with the description error of 9.28%. All moment invariants can be reliable used for object recognition in images with up to four times lower resolution as in original image

Multiscale Edge Detector

The purpose of this paper is to introduce an novel edge detector using directional wavelet transform and signal registration. Directional wavelet transform decomposes an image into four-dimensional space which augments the image by the scale and directional information. We show the directional information significantly improves edge detection in noisy images in comparison with the classical multiscale edge detector, and enables more reliable edge tracing along an object boundary. Edge detection is based on the modulus maximum principle with an adaptive threshold whose value is calculated via maximum entropy measure. Edge tracing is done with signal alignment between neighbour edges

A Spatial Filter for Dark Regions Detection

This paper presents a new spatial filter for detection of dark objects in images with uneven illumination or where object grey-levels vary with respect to local surroundings. Theoretical filter foundations, properties and guidelines for its application are described in detail. This parametric, symmetric, and rotation invariant filter adapts by varying its radius, and, consequently, enables scale-space image analyzing. Filter executes very fast also by big radii, because majority of filter mask elements are zero. Preliminary results are very promising. Proposed filter outperforms related methods like thresholding and local standard deviation in specific images, and, therefore, it is efficient supplement for these methods. With two integrated post-processing methods, this filter successfully deals also with image noise. Consequently, introduced filter can be legitimate reckoned among basic low-level digital image processing tools

Edge Relaxation in Images using Directional Hierarchical Image Decomposition

This paper presents a new hierarchical method of the edge relaxation by using an edge confidence measure. Proposed method is an adaptive and based on directional hierarchical image decomposition and an edge connecting algorithm. It is shown that such combination has low sensitivity to noise, while it is highly robust to outliers, and provide a quality edge connection mechanism