E.R. Davies

A modified Hough scheme for general circle location

Abstract This paper analyses how the Hough transform can be used for the simultaneous detection of circles with a variety of radii. It is found that this is possible using just one instead of a large number of planes in parameter space. A further small amount of computation is necessary to ascertain the radii of circles thus located. This result should save computation in automated inspection and other applications.

Finding ellipses using the generalised Hough transform

Abstract This paper studies how ellipses may be detected using the Hough transform. Early work by Tsuji and Matsumoto permitted ellipses to be detected via pairs of features, e.g. pairs of anti-parallel edge points — an approach that is computation intensive in cluttered images. Alternatively, ellipses may be detected with the generalised Hough transform, if several planes in parameter space can be used to cope with varying object orientation. Here we show that a single plane in parameter space can be employed, with a consequent gain in efficiency. The method is especially suitable for ellipses of low eccentricity.

A high speed algorithm for circular object location

Abstract This paper analyses how the Hough transform approach to circular object location could be speeded up by at least an order of magnitude — e.g. for certain low-cost automated inspection applications. Image sampling is found to be required and an effective but simply implemented strategy is devised which achieves speedup factors of up to ∼25 with real images. The algorithm is highly robust against shape defects and partial occlusions, and in addition robustness is easily monitored in practical situations.

Image space transforms for detecting straight edges in industrial images

Abstract This paper describes a computationally efficient technique, based on the Hough transform, for the detection of straight edges in industrial images. Accuracy is improved by use of the sub-image method, which also permits sensitivity to be optimised for lines of various lengths and curvatures. The approach permits objects to be located within 1 pixel and orientated within ∼1°.

On the noise suppression and image enhancement characteristics of the median, truncated median and mode filters

Abstract This paper examines image filtering operations based on the local intensity histogram. While the mode filter should be useful for image enhancement, it is difficult to implement in a small neighbourhood with sparce intensity distribution, and the ‘truncated median’ filter is posed as an alternative. The latter is easy to implement, and is less prone to error in the vicinity of edges. In practice it is found to be highly effective both for image enhancement and for introducing a certain amount of noise suppression.

Image processing for the food industry

Image processing methodology: images and image processing shape analysis feature detection and object location texture three-dimensional processing pattern recognition. Application to food production: inspection and inspection procedures inspection of baked products cereal grain inspection X-ray inspection image processing in agriculture vision for fish and meat processing system design considerations food processing for the millennium.

A new framework for analysing the properties of the generalised Hough transform

Abstract Certain problems are identified relating to optimal use of the generalised Hough transform for object detection. A new framework is developed which permits these problems to be tackled systematically. It is found that the transform is not a simple matched filter and that it has sub-optimal signal detection capability; however, sensitivity is improved by gradient weighting of points in parameter space, when it becomes proportional to image contrast.

The effect of noise on edge orientation computations

Abstract This paper studies the effect of image noise on edge orientation computations. It is found that noise affects estimation of edge orientation in a complex way, but this is simplified for those ‘circular’ operators which act in a strictly vectorial manner. In that case the distribution of edge orientations is closely gaussian for gaussian image noise. These results have important consequences for object location schemes based on the generalised Hough transform - especially when noise is high or contrast is low. Suprisingly, the consequences are much less serious with impulse noise.

Locating objects from their point features using an optimised Hough-like accumulation technique

Abstract For some years the standard method for locating objects from their salient features has been the maximal clique approach to graph matching. Recently, it was found how the Hough transform could be applied to this purpose, with the accompanying advantage of considerably speeding up the computation. This paper studies the accuracy available with this method, and suggests means for optimising accuracy and minimising computation.

Design of efficient line segment detectors for cereal grain inspection

This paper shows how line segment detection can be achieved with a minimum of computation if two masks embodying a vectorial design strategy are employed. To achieve the greatest combination of accuracy and speed, a two-stage procedure was used with the vectorial operator backed up by a template matching technique. Experimental tests verified the effectiveness of this type of line segment detector for locating insects in cereal grain images: in a set of 60 images containing 150 insects, there were no false positives and the only false negative arose from two insects which were in contact. The approach should be useful in a good many other areas ranging from industrial inspection to document processing and remote sensing--and indeed, in the many applications where line segments have to be located or thin lines have to be tracked.